get_signals

get_signals is the AdCP task for discovering audience and contextual signals. Search by natural language or signal reference, filter by platform and CPM, and get real-time deployment status with activation keys.

Task: Discover signals based on description, with details about where they are deployed.

Response Time: ~60 seconds (inference/RAG with back-end systems)

Request Schema: https://adcontextprotocol.org/schemas/v3/signals/get-signals-request.json Response Schema: https://adcontextprotocol.org/schemas/v3/signals/get-signals-response.json

The get_signals task returns both signal metadata and real-time deployment status across platforms, allowing agents to understand availability and guide the activation process.

Request Parameters

Parameter	Type	Required	Description
discovery_mode	string	No (v3.1+ SHOULD include)	"brief" (default) or "wholesale". "brief": existing behavior — signal_spec, signal_refs, or deprecated signal_ids is required, agent performs inference/RAG. "wholesale": raw wholesale signals feed enumeration — signal_spec, signal_refs, and signal_ids MUST NOT be provided; agent returns its full priced signals feed, paginated, scoped by filters / account / destinations / countries when present. Timing semantics: "wholesale" is a wholesale signals feed read — agents SHOULD respond synchronously and MUST NOT route through the async/Submitted arm; partial completion uses incomplete[]. Agents receiving requests from pre-v3.1 clients without discovery_mode MUST default to "brief". Probe support via get_adcp_capabilities (signals.discovery_modes).
signal_spec	string	Conditional	Natural language description of the desired signals. Required when discovery_mode is "brief" and both signal_refs and legacy signal_ids are absent. MUST NOT be provided when discovery_mode is "wholesale".
signal_refs	SignalRef[]	Conditional	Specific signals to look up by reference. Required when discovery_mode is "brief" and signal_spec is absent. MUST NOT be provided when discovery_mode is "wholesale".
signal_ids	SignalID[]	Conditional	Deprecated. Use signal_refs instead. Legacy specific-signal lookup for older clients. MUST NOT be provided when discovery_mode is "wholesale".
account	AccountRef	No	Account for this request. When provided, the signals agent returns per-account pricing options if configured. In wholesale mode, this is the rate-card scope; when omitted the agent returns default rate-card pricing or omits pricing_options entirely.
destinations	Destination[]	No	Filter signals to those activatable on specific agents/platforms. When omitted, returns all signals available on the current agent. See Destination Object below.
countries	string[]	No	Countries where signals will be used (ISO 3166-1 alpha-2 codes)
filters	Filters	No	Filters to refine results (see Filters Object below). In wholesale mode, filters constrain the enumerated signals feed (e.g., filters.data_providers: ["acme-data"] returns only that provider’s signals).
fields	string[]	No	Specific signal fields to include in the response, aligned with get_products.fields. Required identity and activation fields are always included when required by the response schema. Use for progressive disclosure of rich definition metadata such as taxonomy, data_sources, methodology, segmentation_criteria, criteria_url, onboarder, modeling, countries, consent_basis, restricted_attributes, policy_categories, art9_basis, and data_subject_rights. Agents SHOULD honor requested fields for exact lookup, refinement, small custom-signal result sets, and private/source-native signals when available. fields is a projection request, not an entitlement grant; agents MAY redact requested definition fields unless the caller is authorized for the underlying lineage, methodology, and rights-routing metadata. When consent_basis or art9_basis is projected for another provider’s signal, the value remains provider-declared signal-definition posture; sellers and federating agents MUST NOT substitute their own processing basis. Broad discovery and wholesale pages MAY still return compact pointers to provider-published definitions or disclosure URLs.
if_wholesale_feed_version	string	No	Opaque wholesale_feed_version token from a prior get_signals response. When provided, the agent compares against its current wholesale signals feed version for the caller’s cache_scope and MAY return unchanged: true (with signals omitted) if nothing has changed. See Wholesale feed versioning and Cache layering.
if_pricing_version	string	No	Opaque pricing_version token from a prior response. MUST only be sent together with if_wholesale_feed_version. Evaluation order: if_wholesale_feed_version mismatch → full payload; if_wholesale_feed_version matches but if_pricing_version mismatches → full payload (so the caller sees updated pricing_options); both match → agent MAY return unchanged: true. Agents that don’t track pricing separately ignore this.
max_results	number	No	Deprecated. Use pagination.max_results instead. When both are present, pagination.max_results takes precedence. Will be removed in AdCP 4.0.
pagination	object	No	Pagination envelope. pagination.max_results (max: 100, default: 50) controls page size; pagination.cursor (opaque token from previous response) advances pages. Required for wholesale (signals feeds may be large).
push_notification_config	PushNotificationConfig	No	Optional webhook channel for async terminal completion/failure notifications on brief semantic discovery. submitted responses with a task_id remain pollable through get_task_status (legacy tasks/get) whether or not this field is present. If the request includes this field and the agent returns submitted, the agent MUST deliver at least the terminal completion/failure notification to the configured webhook; intermediate progress notifications are MAY. If the agent cannot honor the webhook channel, it MUST reject the request with a structured error instead of silently accepting. Ignored for wholesale; agents MUST NOT route wholesale reads through the Submitted arm because this field is present.

discovery_mode: "wholesale" is an AdCP 3.1+ discovery shape. When the caller cannot assume a 3.1+ signals surface or wholesale support, clients should first call get_adcp_capabilities and confirm that signals.discovery_modes includes "wholesale" before issuing a wholesale request. If the field is absent or lists only "brief", treat the agent as brief-only and use signal_spec, signal_refs, or deprecated signal_ids.

Async discovery

discovery_mode: "brief" MAY return submitted when semantic discovery depends on slow provider queries or review that cannot finish before the initial response. Polling get_task_status (legacy tasks/get) with task_id is always valid. If push_notification_config is present and the agent returns submitted, the agent also sends at least the terminal completion/failure notification to that webhook; intermediate progress notifications are optional. discovery_mode: "wholesale" remains a synchronous feed read and reports partial completion via incomplete[], not submitted.

Destination Object

Each deployment target uses a type field to discriminate between platform-based and agent-based deployments:

Parameter	Type	Required	Description
type	string	Yes	Discriminator: “platform” for DSPs, “agent” for sales agents
platform	string	Conditional*	Platform identifier (e.g., ‘the-trade-desk’, ‘amazon-dsp’). Required when type=“platform”
agent_url	string (URI)	Conditional*	URL identifying the sales agent. Required when type=“agent”
account	string	No	Account identifier on the platform or agent

*platform is required when type="platform", agent_url is required when type="agent".

Destination filtering: Signals are returned if they are available on any of the requested destinations (OR semantics). Destinations where a signal is not available are omitted from that signal’s response deployments array. A PARTIAL_COVERAGE warning may be included when some destinations don’t support the signal.

Activation Keys: If the authenticated caller has access to any of the destinations in the request, the signal agent will include activation_key fields in the response for those deployments (when is_live: true).

Permission Model: The signal agent determines key inclusion based on the caller’s authentication and authorization. For example:

• A sales agent receives keys for deployments matching its agent_url
• A buyer with credentials for multiple DSP platforms receives keys for all those deployments
• Access is determined by the signal agent’s permission system, not by flags in the request

Filters Object

Parameter	Type	Required	Description
catalog_types	string[]	No	Filter by catalog type (“marketplace”, “custom”, “owned”)
data_providers	string[]	No	Filter by specific data providers
max_cpm	number	No	Maximum CPM price filter. Excludes signals where all CPM-based pricing options exceed this value. Signals without CPM-based pricing options are not affected by this filter.
min_coverage_percentage	number	No	Minimum coverage requirement

catalog_types, the deprecated catalog_signals capability flag, and the deprecated signal_id.source: "catalog" value are legacy wire terms. In new prose, read them as provider-published signal definitions in adagents.json signals[]. Existing 3.x agents may continue to accept or emit them for compatibility, but new callers SHOULD use signal_ref and MUST NOT require signals.features.catalog_signals before using the Signals protocol.

Response Structure

All AdCP responses include:

• message: Human-readable summary of the operation result
• context_id: Session continuity identifier for follow-up requests
• data: Task-specific payload (see Response Data below)

The response structure is identical across protocols, with only the transport wrapper differing:

• MCP: Returns complete response as flat JSON
• A2A: Returns as artifacts with message in text part, data in data part

Response Data

{
  "signals": [
    {
      "signal_ref": {
        "scope": "data_provider",
        "data_provider_domain": "string",
        "signal_id": "string"
      },
      "signal_agent_segment_id": "string",
      "name": "string",
      "description": "string",
      "signal_type": "string",
      "data_provider": "string",
      "coverage_percentage": "number (optional, deprecated)",
      "coverage_forecast": {
        "method": "estimate",
        "forecast_range_unit": "availability",
        "scope": {
          "kind": "inventory",
          "label": "network price-priority inventory"
        },
        "bucket_semantics": "exclusive",
        "bucket_completeness": "partial",
        "points": [
          {
            "label": "not present",
            "dimensions": [
              {
                "kind": "signal",
                "signal_ref": {
                  "scope": "data_provider",
                  "data_provider_domain": "weather-data.example",
                  "signal_id": "weather"
                },
                "signal_value": null,
                "presence": "absent"
              }
            ],
            "metrics": {
              "impressions": { "mid": 280000 },
              "coverage_rate": { "mid": 0.28 }
            }
          },
          {
            "label": "hot",
            "dimensions": [
              {
                "kind": "signal",
                "signal_ref": {
                  "scope": "data_provider",
                  "data_provider_domain": "weather-data.example",
                  "signal_id": "weather"
                },
                "signal_value": "hot",
                "presence": "present"
              }
            ],
            "metrics": {
              "impressions": { "mid": 180000 },
              "coverage_rate": { "mid": 0.18 }
            }
          }
        ]
      },
      "deployments": [
        {
          "type": "agent",
          "agent_url": "string",
          "account": "string",
          "is_live": "boolean",
          "activation_key": {
            "type": "segment_id",
            "segment_id": "string"
          },
          "estimated_activation_duration_minutes": "number"
        }
      ],
      "pricing_options": [
        {
          "pricing_option_id": "string",
          "model": "cpm | percent_of_media | flat_fee | per_unit | custom",
          "...": "..."
        }
      ]
    }
  ]
}

get_signals is a discovery and availability surface, not a requirement to inline every field from the authoritative signal definition. For broad search results and wholesale feed pages, signal agents SHOULD keep each listing compact and use stable references plus cacheable disclosure pointers for large definition resources. Buyers can dereference provider-published definitions from signal_ref: fetch the provider’s /.well-known/adagents.json (following authoritative_location when present) and select the matching signals[].id. Cache the fetched adagents.json document by resolved authoritative URL plus catalog_etag, HTTP ETag / Last-Modified, or a bounded TTL, then resolve signal_ref.signal_id within that cached document. Use taxonomy.etag only for taxonomy documents that have their own freshness validator. These reuse existing provider-file and taxonomy validators; get_signals does not define a separate signal-definition validator, and clients must not key signal-definition caches by validator alone.

Definition field inclusion

Buyers that need richer review context in the same call can set fields. This uses the same response-projection pattern as get_products.fields, rather than introducing a separate lookup task.

Required identity and activation fields are always included when required by the response schema. Additional values request optional listing fields or rich definition metadata inline, including taxonomy, data_sources, methodology, segmentation_criteria, criteria_url, refresh_cadence, lookback_window, onboarder, modeling, audience_expansion, device_expansion, countries, consent_basis, restricted_attributes, policy_categories, art9_basis, and data_subject_rights.

When consent_basis or art9_basis is projected for another provider’s signal, the value remains provider-declared signal-definition posture. Sellers and federating agents MUST NOT substitute their own processing basis for the provider-declared basis.

Agents SHOULD honor requested fields for exact lookup, refinement, and small custom-signal result sets when the fields are available. For broad discovery and wholesale pages, agents MAY still return compact listings with pointers to provider-published definitions, taxonomy documents, criteria pages, and disclosure URLs when inlining the requested fields would make the page too large. This keeps static signals cacheable through adagents.json while letting custom or brief-specific signals return deeper inline context when useful.

Field Descriptions

• signals: Array of matching signals
  • signal_ref: Canonical signal reference. Use scope: "data_provider" for signals resolved through adagents.json signals[], scope: "signal_source" for source-native signals that are not published in adagents.json signals[], and scope: "product" only in product-contextual responses. New responses SHOULD include this field.
  • signal_id: Deprecated legacy SignalId object. New clients should read signal_ref; during the migration window, older responses may include only signal_id.
  • signal_agent_segment_id: Opaque signal handle issued by this signal source. Use it verbatim for activate_signal; do not treat it as a globally portable signal ID. For package-level signal_targeting_groups, signal_ref is the buy-time identity and this handle is echoed only when the selected product option exposes it as a separate execution handle.
  • name: Human-readable signal name
  • description: Detailed signal description
  • signal_type: Type of signal. One of:
    • marketplace — resold third-party segment (provider authorization verifiable via the provider’s adagents.json)
    • owned — first-party segment derived from data the signal source directly owns
    • custom — source-native segment built on demand from models, composites, or buyer inputs (not attributable to a standing upstream provider)
  • data_provider: Human-readable source name when applicable. For scope: "data_provider" signals this is the data provider; for scope: "signal_source" signals it may identify the signal source or proprietary origin.
  • coverage_percentage: Optional deprecated legacy scalar percentage of audience coverage. Use only as a fallback for clients that do not consume coverage_forecast. When coverage_forecast is present, coverage_forecast is authoritative for signal-level discovery and this scalar is fallback-only. If coverage_forecast includes an absent bucket over the same denominator, coverage_percentage should align with 100 * (1 - absent coverage_rate.mid).
  • coverage_forecast: Optional forecast-shaped availability guidance for the signal. scope declares the denominator, bucket_semantics declares whether returned value buckets are exclusive or overlapping, and bucket_completeness declares whether the returned buckets are a full denominator partition or a partial histogram. Each point can use a kind: "signal" dimension with canonical signal_ref, presence: "present" for any present value, presence: "present" plus signal_value for a specific value bucket, or presence: "absent" plus signal_value: null for the not-present bucket. metrics.coverage_rate is a 0.0-1.0 fraction of the declared scope.
  • deployments: Array of destination deployments
    • agent_url: URL identifying the destination agent
    • account: Account identifier if applicable
    • is_live: Whether signal is currently active on this deployment
    • activation_key: The key to use for targeting (see Activation Key below). Only present when is_live=true and the authenticated caller has access to this deployment.
    • estimated_activation_duration_minutes: Time to activate if not live
  • pricing_options: Array of pricing options for this signal when it has an incremental price. Pass the selected pricing_option_id in report_usage or package-level signal_targeting_groups for billing verification. Omitted when pricing is unavailable to the caller, bundled into the destination product, or has no incremental cost.
    • pricing_option_id: Unique identifier for this pricing option
    • model: Pricing model — cpm, percent_of_media, flat_fee, per_unit, or custom
    • model: "cpm" — cpm (number, cost per thousand impressions), currency (ISO 4217)
    • model: "percent_of_media" — percent (0–100), currency (ISO 4217), max_cpm (optional CPM cap: effective charge = min(percent × media_spend_per_mille, max_cpm))
    • model: "flat_fee" — amount (fixed charge), currency (ISO 4217), period (monthly, quarterly, annual, or campaign)
    • model: "per_unit" — unit (what is counted), unit_price (cost per one unit), currency (ISO 4217)
    • model: "custom" — description (human-readable), metadata (structured parameters), currency (optional). Escape hatch for performance kickers, tiered volume, hybrid formulas, or any construct the standard models cannot express. Buyers SHOULD route custom pricing through operator review before commitment.

Select the pricing option that matches your billing model. For direct signal activation, pass its pricing_option_id in report_usage for billing verification; for seller-offered signals selected on a media product, pass it in package-level targeting_overlay.signal_targeting_groups.groups[].signals[].pricing_option_id. If a signal offers multiple models (e.g., CPM and flat fee), choose based on your expected delivery volume and campaign structure.

Response Metadata

Field	Type	Description
wholesale_feed_version	string	Opaque token representing the version of the wholesale signals feed state used to compose this response. Treat as opaque — no format, no ordering, no inspection. Returned on every response by agents implementing conditional-fetch. See Wholesale feed versioning.
pricing_version	string	Optional finer-grained token. Changes when prices move while wholesale_feed_version only changes for structure/metadata. Agents not separating these MAY omit pricing_version.
cache_scope	string	"public" or "account". REQUIRED on every response (schema-enforced — the safety property of the two-layer cache depends on it). When the request had no account, MUST be "public". When the request had account, the agent declares either "public" (account prices off the rate card — caller dedupes) or "account" (account-specific overrides). See Cache layering.
unchanged	boolean	Present and true ONLY when the request carried if_wholesale_feed_version (and/or if_pricing_version) matching the agent’s current version for the caller’s cache_scope, in which case signals[] MUST be omitted; wholesale_feed_version, cache_scope, and pricing_version (when used) MUST still be echoed. Agents MUST NOT emit unchanged: false — absence of the field IS the “response carries signals” signal (one shape per state). Callers receiving unchanged: true MUST NOT mutate their local wholesale signals mirror.
incomplete	IncompleteEntry[]	Declares what the agent could not finish within the caller’s time_budget or due to internal limits. See Incomplete array.
pagination	PaginationResponse	has_more, cursor, optional total_count. Required for wholesale mode.

Activation Key Object

The activation key represents how to use the signal on a deployment target. It can be either a segment ID or a key-value pair:

Segment ID format:

{
  "type": "segment_id",
  "segment_id": "ttd_segment_12345"
}

Key-Value format:

{
  "type": "key_value",
  "key": "audience_segment",
  "value": "luxury_auto_intenders"
}

Protocol-Specific Examples

The AdCP payload is identical across protocols. Only the request/response wrapper differs.

MCP Request - Sales Agent Requesting Signals

A sales agent querying for signals. Because the authenticated caller is wonderstruck.salesagents.com, the signal agent will include activation keys in the response:

{
  "tool": "get_signals",
  "arguments": {
    "signal_spec": "High-income households interested in luxury goods",
    "destinations": [
      {
        "type": "agent",
        "agent_url": "https://wonderstruck.salesagents.com"
      }
    ],
    "countries": ["US"],
    "filters": {
      "max_cpm": 5.0,
      "catalog_types": ["marketplace"]
    },
    "pagination": {
      "max_results": 5
    }
  }
}

MCP Response - With Activation Key

Because the authenticated caller matches the deployment target, the response includes the activation key:

{
  "$schema": "/schemas/signals/get-signals-response.json",
  "status": "completed",
  "message": "Found 1 luxury segment matching your criteria. Already activated for your sales agent.",
  "context_id": "ctx-signals-123",
  "cache_scope": "public",
  "wholesale_feed_version": "sig_v2026-05-18T08:00:00Z-public-rev12",
  "signals": [
    {
      "signal_ref": {
        "scope": "data_provider",
        "data_provider_domain": "pinnacle-data.example",
        "signal_id": "luxury_auto_intenders"
      },
      "signal_agent_segment_id": "luxury_auto_intenders",
      "name": "Luxury Automotive Intenders",
      "description": "High-income individuals researching luxury vehicles",
      "signal_type": "marketplace",
      "data_provider": "Pinnacle Data",
      "coverage_forecast": {
        "method": "estimate",
        "forecast_range_unit": "availability",
        "scope": {
          "kind": "inventory",
          "label": "eligible destination inventory"
        },
        "bucket_semantics": "exclusive",
        "bucket_completeness": "partial",
        "points": [
          {
            "label": "present",
            "dimensions": [
              {
                "kind": "signal",
                "signal_ref": {
                  "scope": "data_provider",
                  "data_provider_domain": "pinnacle-data.example",
                  "signal_id": "luxury_auto_intenders"
                },
                "presence": "present"
              }
            ],
            "metrics": {
              "coverage_rate": { "mid": 0.12 }
            }
          }
        ]
      },
      "deployments": [
        {
          "type": "agent",
          "agent_url": "https://wonderstruck.salesagents.com",
          "is_live": true,
          "activation_key": {
            "type": "key_value",
            "key": "audience_segment",
            "value": "luxury_auto_intenders_v2"
          }
        }
      ],
      "pricing_options": [
        {
          "pricing_option_id": "po_cpm_usd",
          "model": "cpm",
          "cpm": 3.50,
          "currency": "USD"
        }
      ]
    }
  ]
}

MCP Response - Multiple Pricing Options

Some signals offer multiple pricing models. The buyer selects one and passes its pricing_option_id in report_usage for direct signal usage or in package-level signal_targeting_groups when the signal is selected on a media buy:

{
  "$schema": "/schemas/signals/get-signals-response.json",
  "status": "completed",
  "message": "Found 1 segment matching your criteria. Three pricing options are available: CPM at $3.50, 15% of media spend, or $5,000/month flat fee.",
  "context_id": "ctx-signals-456",
  "cache_scope": "public",
  "wholesale_feed_version": "sig_v2026-05-18T08:00:00Z-public-rev12",
  "signals": [
    {
      "signal_ref": {
        "scope": "data_provider",
        "data_provider_domain": "acmedata.com",
        "signal_id": "eco_conscious_shoppers"
      },
      "signal_agent_segment_id": "eco_conscious_shoppers",
      "name": "Eco-Conscious Shoppers",
      "description": "Users with demonstrated interest in sustainable and eco-friendly products",
      "signal_type": "marketplace",
      "data_provider": "Acme Data",
      "coverage_percentage": 18,
      "deployments": [
        {
          "type": "agent",
          "agent_url": "https://wonderstruck.salesagents.com",
          "is_live": true,
          "activation_key": {
            "type": "segment_id",
            "segment_id": "eco_seg_789"
          }
        }
      ],
      "pricing_options": [
        {
          "pricing_option_id": "po_eco_cpm",
          "model": "cpm",
          "cpm": 3.50,
          "currency": "USD"
        },
        {
          "pricing_option_id": "po_eco_pom",
          "model": "percent_of_media",
          "percent": 15,
          "max_cpm": 1.50,
          "currency": "USD"
        },
        {
          "pricing_option_id": "po_eco_flat",
          "model": "flat_fee",
          "amount": 5000,
          "period": "monthly",
          "currency": "USD"
        }
      ]
    }
  ]
}

MCP Request - Buyer Querying Multiple DSP Platforms

A buyer checking availability across multiple DSP platforms:

{
  "tool": "get_signals",
  "arguments": {
    "signal_spec": "High-income households interested in luxury goods",
    "destinations": [
      {
        "type": "platform",
        "platform": "the-trade-desk",
        "account": "agency-123"
      },
      {
        "type": "platform",
        "platform": "amazon-dsp"
      }
    ],
    "countries": ["US"],
    "filters": {
      "max_cpm": 5.0,
      "catalog_types": ["marketplace"]
    },
    "pagination": {
      "max_results": 5
    }
  }
}

MCP Response - Buyer With Multi-Platform Access

A buyer with credentials for both The Trade Desk and Amazon DSP receives keys for both platforms:

{
  "$schema": "/schemas/signals/get-signals-response.json",
  "status": "completed",
  "message": "Found 1 luxury segment matching your criteria. Already activated on The Trade Desk, pending activation on Amazon DSP.",
  "context_id": "ctx-signals-123",
  "cache_scope": "public",
  "wholesale_feed_version": "sig_v2026-05-18T08:00:00Z-public-rev12",
  "signals": [
    {
      "signal_ref": {
        "scope": "data_provider",
        "data_provider_domain": "experian.com",
        "signal_id": "luxury_auto_intenders"
      },
      "signal_agent_segment_id": "luxury_auto_intenders",
      "name": "Luxury Automotive Intenders",
      "description": "High-income individuals researching luxury vehicles",
      "signal_type": "marketplace",
      "data_provider": "Experian",
      "coverage_percentage": 12,
      "deployments": [
        {
          "type": "platform",
          "platform": "the-trade-desk",
          "account": "agency-123",
          "is_live": true,
          "activation_key": {
            "type": "segment_id",
            "segment_id": "ttd_agency123_exp_lux_auto"
          }
        },
        {
          "type": "platform",
          "platform": "amazon-dsp",
          "is_live": false,
          "estimated_activation_duration_minutes": 60
        }
      ],
      "pricing_options": [
        {
          "pricing_option_id": "po_cpm_usd",
          "model": "cpm",
          "cpm": 3.50,
          "currency": "USD"
        }
      ]
    }
  ]
}

A2A Request

Natural Language Invocation

await a2a.send({
  message: {
    parts: [{
      kind: "text",
      text: "Find me signals for high-income households interested in luxury goods that can be deployed on The Trade Desk and Amazon DSP in the US, with a maximum CPM of $5.00."
    }]
  }
});

Explicit Skill Invocation

await a2a.send({
  message: {
    parts: [{
      kind: "data",
      data: {
        skill: "get_signals",
        parameters: {
          signal_spec: "High-income households interested in luxury goods",
          destinations: [
            {
              type: "agent",
              agent_url: "https://thetradedesk.com",
              account: "agency-123"
            },
            {
              type: "agent",
              agent_url: "https://advertising.amazon.com/dsp"
            }
          ],
          countries: ["US"],
          filters: {
            max_cpm: 5.0,
            catalog_types: ["marketplace"]
          },
          pagination: {
            max_results: 5
          }
        }
      }
    }]
  }
});

A2A Response

A2A returns results as artifacts with the same data structure:

{
  "artifacts": [{
      "artifactId": "artifact-signal-discovery-def456",
      "name": "signal_discovery_result",
      "parts": [
        {
          "kind": "text",
          "text": "Found 1 luxury segment matching your criteria. Available on The Trade Desk, pending activation on Amazon DSP."
        },
        {
          "kind": "data",
          "data": {
            "context_id": "ctx-signals-123",
            "signals": [
              {
                "signal_ref": {
                  "scope": "data_provider",
                  "data_provider_domain": "experian.com",
                  "signal_id": "luxury_auto_intenders"
                },
                "signal_agent_segment_id": "luxury_auto_intenders",
                "name": "Luxury Automotive Intenders",
                "description": "High-income individuals researching luxury vehicles",
                "signal_type": "marketplace",
                "data_provider": "Experian",
                "coverage_percentage": 12,
                "deployments": [
                  {
                    "type": "agent",
                    "agent_url": "https://thetradedesk.com",
                    "account": "agency-123",
                    "is_live": true
                  },
                  {
                    "type": "agent",
                    "agent_url": "https://advertising.amazon.com/dsp",
                    "is_live": false,
                    "estimated_activation_duration_minutes": 60
                  }
                ],
                "pricing_options": [
                  {
                    "pricing_option_id": "po_cpm_usd",
                    "model": "cpm",
                    "cpm": 3.50,
                    "currency": "USD"
                  }
                ]
              }
            ]
          }
        }
      ]
    }]
}

Protocol Transport

• MCP: Direct tool call with arguments, returns complete response as flat JSON
• A2A: Skill invocation with input, returns structured artifacts with message and data separated
• Data Consistency: Both protocols contain identical AdCP data structures and version information

Scenarios

All Platforms Discovery

Discover all available deployments across platforms:

{
  "$schema": "/schemas/signals/get-signals-request.json",
  "signal_spec": "Contextual segments for luxury automotive content",
  "destinations": [
    { "type": "platform", "platform": "index-exchange", "account": "agency-123-ix" },
    { "type": "platform", "platform": "openx" },
    { "type": "platform", "platform": "pubmatic", "account": "brand-456-pm" }
  ],
  "countries": ["US"],
  "filters": {
    "data_providers": ["Peer39"],
    "catalog_types": ["marketplace"]
  }
}

Response

Message: “Found luxury automotive contextual segment from Peer39 with 15% coverage. Live on Index Exchange and OpenX, pending activation on Pubmatic.”

Payload:

{
  "$schema": "/schemas/signals/get-signals-response.json",
  "status": "completed",
  "cache_scope": "public",
  "wholesale_feed_version": "sig_v2026-05-18T08:00:00Z-public-rev12",
  "signals": [{
    "signal_ref": {
      "scope": "data_provider",
      "data_provider_domain": "peer39.com",
      "signal_id": "peer39_luxury_auto"
    },
    "signal_agent_segment_id": "peer39_luxury_auto",
    "name": "Luxury Automotive Context",
    "description": "Pages with luxury automotive content and high viewability",
    "signal_type": "marketplace",
    "data_provider": "Peer39",
    "coverage_percentage": 15,
    "deployments": [
      {
        "type": "platform",
        "platform": "index-exchange",
        "account": "agency-123-ix",
        "is_live": true,
        "activation_key": {
          "type": "segment_id",
          "segment_id": "ix_agency123_peer39_lux_auto"
        }
      },
      {
        "type": "platform",
        "platform": "index-exchange",
        "is_live": true,
        "activation_key": {
          "type": "segment_id",
          "segment_id": "ix_peer39_luxury_auto_gen"
        }
      },
      {
        "type": "platform",
        "platform": "openx",
        "is_live": true,
        "activation_key": {
          "type": "segment_id",
          "segment_id": "ox_peer39_lux_auto_456"
        }
      },
      {
        "type": "platform",
        "platform": "pubmatic",
        "account": "brand-456-pm",
        "is_live": false,
        "estimated_activation_duration_minutes": 60
      }
    ],
    "pricing_options": [
      {
        "pricing_option_id": "po_cpm_usd",
        "model": "cpm",
        "cpm": 2.50,
        "currency": "USD"
      }
    ]
  }]
}

Response Fields

• context_id (string): Context identifier for session persistence
• signals (array): Array of matching signals
  • signal_agent_segment_id (string): Opaque signal handle issued by this signal source. Use it verbatim for activate_signal; do not treat it as a globally portable signal ID. For media-buy signal groups, signal_ref is the buy-time identity and this handle is echoed only when the selected product option exposes it as a separate execution handle.
  • name (string): Human-readable signal name
  • description (string): Detailed signal description
  • signal_type (string): marketplace (resold third-party), owned (signal source first-party data), or custom (source-native segment built on demand)
  • data_provider (string, optional): Human-readable source/provider name when applicable
  • coverage_percentage (number, optional, deprecated): Legacy scalar estimated reach percentage. Use only as a fallback when coverage_forecast is absent or unsupported by the client.
  • coverage_forecast (object, optional): Forecast-shaped coverage breakdown with an explicit denominator and availability points. Use presence: "present" with omitted signal_value for an aggregate “any present value” bucket; add signal_value only when the row is for a specific value. Use bucket_semantics: "exclusive" when returned buckets do not overlap, or "overlapping" when multi-value signals can make returned rates sum above 1.0. Use bucket_completeness: "complete" only when the returned buckets cover the declared denominator; otherwise use "partial" and buyers must treat omitted share as undisclosed, other, or unsupported buckets.

Use coverage_forecast as the authoritative field for signal-first discovery: “how much inventory has this signal or its values?” Use product or proposal forecasts with kind: "signal" dimensions as the authoritative surface for product-specific planning: “how does this signal restrict this product’s baseline availability?” Bucket-specific filtering is client-side for now; request filters such as min_coverage_percentage still use the legacy scalar.

• deployments (array): Platform-specific deployment information
  • platform (string): Target platform name
  • account (string, nullable): Specific account if account-specific
  • is_live (boolean): Whether signal is currently active
  • activation_key (object): The key to use for targeting. Only present when is_live=true and the caller has access. See Activation Key Object above.
  • estimated_activation_duration_minutes (number, optional): Time to activate if not live
• pricing_options (array): Array of pricing options available for this signal when it has an incremental price. Select one and pass its pricing_option_id in report_usage or package-level signal_targeting_groups.
  • pricing_option_id (string): Unique identifier for this pricing option
  • model (string): Pricing model — cpm, percent_of_media, flat_fee, per_unit, or custom

Error Codes

Discovery Errors

• REFERENCE_NOT_FOUND: Referenced signal_agent_segment_id doesn’t exist, OR a private signal agent is not visible to this account. The same code is returned whether the resource exists but is unauthorized or truly does not exist — sellers MUST NOT distinguish the two (see error.field to identify which typed parameter failed to resolve). See the uniform-response MUST in error-handling.mdx.
• AGENT_ACCESS_DENIED: Authenticated agent’s credentials did not authorize access to this signal agent

Discovery Warnings

• PRICING_UNAVAILABLE: Pricing data temporarily unavailable for one or more platforms
• PARTIAL_COVERAGE: Some requested platforms don’t support this signal type
• STALE_DATA: Some signal metadata may be outdated due to provider refresh delays

Usage Notes

1. Authentication-Based Keys: Activation keys are only returned when the authenticated caller matches one of the deployment targets
2. Permission Security: The signal agent determines key inclusion based on caller identity, not request flags
3. Deployment Status: Check is_live to determine if activation is needed
4. Multiple Deployments: Query multiple deployment targets to check availability across platforms
5. Activation Required: If is_live is false, use the activate_signal task
6. The message field provides a quick summary of the most relevant findings

Seller-offered signals on media products

A sales agent that owns or is authorized to apply targeting signals exposes buy-time product eligibility through get_products. It can also expose a broader cross-product signal feed through get_signals when buyers need discovery or activation before package selection:

• get_products declares whether a product has package-level signal targeting. products[].included_signals describes non-selectable signals already bundled into or planned into a product. Inline products[].signal_targeting_options can carry a product-specific menu, price, activation handle, default, grouping hint, or brief/refine-selected subset; wholesale products can omit inline options and use get_signals as the selectable signal feed.
• get_signals optionally returns cross-product signal metadata, including signal_ref, signal_agent_segment_id, value metadata, and any default or account-scoped pricing_options.
• create_media_buy selects the signal for a package in packages[].targeting_overlay.signal_targeting_groups, carrying the selected signal pricing_option_id, signal_ref, and any separate seller execution handle when required.

This gives buyers a product-first buy-time eligibility path without forcing products to duplicate a large wholesale signal feed: use the selected product’s inline signal_targeting_options when present and signal_targeting_rules to determine what may be applied on that package. When a product omits inline options but allows signal targeting, use get_signals for candidate discovery and activation metadata, then use get_products.filters.signal_targeting or a product-specific get_products query to confirm the candidate set is selectable and jointly composable for the intended product before calling create_media_buy. When both surfaces include pricing, the product-scoped signal_targeting_options[].pricing_options price is authoritative for that product. For product-local signals exposed on both surfaces, signal_ref.signal_id in the product option MUST match the seller’s get_signals.signals[].signal_ref.signal_id for the same signal. included_signals is descriptive only and does not make a signal selectable.

For media-buy product targeting, product options and package groups use the same signal_ref identity: scope: "product" for product-local signal options, scope: "data_provider" with data_provider_domain for signals defined in published adagents.json signals[], or scope: "signal_source" for source-native custom signals that are not published in adagents.json signals[]. When a provider-published signal is selected, buyers can verify the seller’s authorization by checking the provider’s adagents.json authorized_agents rules for the signal id or tags. The older Signals Protocol signal_id.source shape is deprecated and retained only for backwards compatibility.

Wholesale signals feed

When a consumer (storefront, federated marketplace, registry) needs to mirror a signals agent’s full priced signals feed, set discovery_mode: "wholesale" and omit signal_spec / signal_refs / deprecated signal_ids. When the caller cannot assume a 3.1+ signals surface or wholesale support, first check get_adcp_capabilities.signals.discovery_modes for "wholesale". Wholesale enumeration is symmetric with get_products buying_mode: "wholesale": synchronous, paginated, firm-priced, with partial completion declared via incomplete[].

Request

{
  "$schema": "/schemas/signals/get-signals-request.json",
  "discovery_mode": "wholesale",
  "account": { "account_id": "acct_123" },
  "filters": {
    "catalog_types": ["marketplace", "owned"],
    "data_providers": ["acme-data", "nova-insights"]
  },
  "pagination": { "max_results": 50 }
}

Response

{
  "$schema": "/schemas/signals/get-signals-response.json",
  "status": "completed",
  "message": "Returning 50 of 312 signals in wholesale mode.",
  "context_id": "ctx-wholesale-001",
  "cache_scope": "public",
  "wholesale_feed_version": "v2026-05-18T08:00:00Z-acme-rev412",
  "signals": [
    {
      "signal_ref": { "scope": "data_provider", "data_provider_domain": "acme-data.com", "signal_id": "luxury_auto_intenders" },
      "signal_agent_segment_id": "sigagent_seg_4421",
      "name": "Luxury Auto Intenders",
      "description": "Households researching premium vehicles in the last 30 days.",
      "signal_type": "marketplace",
      "data_provider": "Acme Data",
      "coverage_percentage": 18.4,
      "deployments": [
        { "type": "platform", "platform": "the-trade-desk", "is_live": true,
          "activation_key": { "type": "segment_id", "segment_id": "ttd_seg_99821" } }
      ],
      "pricing_options": [
        { "pricing_option_id": "po_cpm_1", "model": "cpm", "cpm": 2.50, "currency": "USD" }
      ]
    }
  ],
  "pagination": { "has_more": true, "cursor": "eyJvIjo1MH0=", "total_count": 312 }
}

Authorization and provenance

Marketplace signals (signal_type: "marketplace") remain attributable to their upstream data provider. Wholesale enumeration does not collapse provenance:

• Each marketplace signal carries data_provider.
• Consumers SHOULD verify provider authorization via the provider’s adagents.json — the signals agent’s URL must appear in the provider’s authorization list for that signal class.
• Storefronts and registries MAY use wholesale enumeration plus adagents.json cross-reference to materialize a data-publisher signal view: for each known data provider, the set of signals available via authorized signals agents, with prices.

Pricing

pricing_options[] MUST be populated for authenticated callers when the agent has firm standalone signal prices to declare. When account is omitted, the agent returns default rate-card pricing or omits pricing_options[] (in which case the caller MUST re-query with account or use product-scoped pricing from get_products before composing). Unauthenticated callers MAY receive signal metadata without pricing. Signals bundled into a media product or carrying no incremental cost MAY omit pricing_options[].

Capability declaration

Signals agents declare wholesale support in get_adcp_capabilities:

{
  "signals": {
    "discovery_modes": ["brief", "wholesale"]
  }
}

Agents not declaring "wholesale" MAY return INVALID_REQUEST for wholesale calls. Because wholesale discovery is AdCP 3.1+, the capability declaration is the canonical support signal when a caller cannot assume a 3.1+ signals surface or wholesale support; callers SHOULD probe before issuing wholesale requests.

Wholesale feed versioning

Even with wholesale enumeration, a consumer mirroring an agent’s signals feed would re-fetch every paginated page on each poll just to detect changes. To avoid that, get_signals supports an opaque wholesale_feed_version token returned on every response. Pass it back via if_wholesale_feed_version on a subsequent call and the agent MAY short-circuit with unchanged: true — no signal payload, no per-page diff.

This is the seller-side wholesale signals feed returned by get_signals. It is not a sync_catalogs feed; sync_catalogs manages buyer-provided campaign input feeds on a seller account.

Unchanged response

Request:

{
  "$schema": "/schemas/signals/get-signals-request.json",
  "discovery_mode": "wholesale",
  "if_wholesale_feed_version": "v2026-05-18T08:00:00Z-acme-rev412"
}

Response (wholesale signals feed unchanged):

{
  "$schema": "/schemas/signals/get-signals-response.json",
  "status": "completed",
  "message": "Wholesale signals feed unchanged since v2026-05-18T08:00:00Z-acme-rev412.",
  "context_id": "ctx-abc-789",
  "unchanged": true,
  "wholesale_feed_version": "v2026-05-18T08:00:00Z-acme-rev412",
  "pricing_version": "v2026-05-18T08:00:00Z-acme-rev412",
  "cache_scope": "public"
}

When unchanged: true, signals[] MUST be omitted and consumers MUST NOT mutate their local wholesale signals mirror.

Wholesale signals feed changed — full payload returned (abbreviated)

test=false

{
  "message": "Returning 312 signals (wholesale feed version advanced).",
  "context_id": "ctx-abc-790",
  "wholesale_feed_version": "v2026-05-18T10:15:00Z-acme-rev415",
  "pricing_version": "v2026-05-18T10:15:00Z-acme-rev415",
  "cache_scope": "public",
  "signals": [
    {
      "signal_ref": { "scope": "data_provider", "data_provider_domain": "acme-data.com", "signal_id": "luxury_auto_intenders" },
      "signal_agent_segment_id": "sigagent_seg_4421",
      "name": "Luxury Auto Intenders",
      "description": "Households researching premium vehicles in the last 30 days.",
      "signal_type": "marketplace",
      "data_provider": "Acme Data",
      "coverage_percentage": 18.4,
      "deployments": [
        { "type": "platform", "platform": "the-trade-desk", "is_live": true,
          "activation_key": { "type": "segment_id", "segment_id": "ttd_seg_99821" } }
      ],
      "pricing_options": [
        { "pricing_option_id": "po_cpm_1", "model": "cpm", "cpm": 2.50, "currency": "USD" }
      ]
    }
  ],
  "pagination": { "has_more": true, "cursor": "eyJvIjo1MH0=", "total_count": 312 }
}

Rules

• Tokens are opaque. No format, no ordering, no inspection.
• A returned wholesale_feed_version is scope-keyed via cache_scope. Callers cache (cache_scope, wholesale_feed_version) pairs alongside the (account, filters, discovery_mode, destinations, countries) tuple used. See Cache layering for the two-layer model.
• pricing_version is an optional finer-grained token: when present, it changes when prices move but wholesale_feed_version changes only when structure/metadata moves. Common for rate-card sweeps that don’t change segment metadata.
• if_pricing_version requires if_wholesale_feed_version. Pricing has no structural baseline of its own. Sending if_pricing_version without if_wholesale_feed_version is a schema-level error. The agent’s evaluation is two-stage: wholesale feed mismatch returns the full payload; wholesale feed match with pricing mismatch also returns the full payload (so the caller sees updated pricing_options); both match → unchanged: true.
• filters canonicalization. Agents MUST treat the filters object as canonicalized before hashing into the wholesale_feed_version keyspace: keys sorted lexicographically, omitted-and-default values treated identically, array values sorted where the filter has set semantics (e.g., catalog_types, data_providers). Callers that pass equivalent-but-differently-shaped filter objects MUST receive the same wholesale_feed_version. Prevents silent stale-mirror bugs from key-order or default-elision differences. Forward-compat default: new filter fields added in 3.x minor versions MUST declare set-vs-sequence semantics; absent an explicit declaration, the rule defaults to set-semantics.
• Pagination interaction. Agents MUST return wholesale_feed_version on every paginated page (not only the first) when they declare wholesale_feed_versioning.supported: true; agents that do not declare versioning SHOULD do the same. If the wholesale feed mutates between pages, the new version surfaces on the next page and the caller MUST restart pagination from cursor: null — the partial pages already received describe a stale version.
• unchanged: true and in-progress pagination. A caller mid-pagination MAY send if_wholesale_feed_version matching the version their pages so far were drawn from. If the agent confirms unchanged: true, the response omits signals[] and pagination envelope entirely; the caller abandons their in-progress walk under that version. Agents MAY NOT use the conditional-fetch short-circuit to skip individual pages within an active pagination — unchanged is feed-versus-cached-version, not per-page.
• Pre-v3.1 agents that ignore if_wholesale_feed_version simply return the full payload — semantically correct, just inefficient.

For pushed change tracking beyond conditional fetch, see specs/wholesale-feed-webhooks.md. Wholesale feed webhooks carry the changed signal payload, pricing payload, removal tombstone, or bulk-change summary; get_signals remains the repair and reconciliation read.

Cache layering

Signals agents publish two notional layers: a public layer (the rate-card / structural view) and per-account overlays (account-specific pricing for premium buyers). The conditional-fetch path is layer-aware via cache_scope.

Two-layer cache.

Layer	Cache key	What’s stored
Public	(agent, discovery_mode, filters, destinations, countries)	wholesale_feed_version_public, the wholesale signals feed payload as seen without an account ref
Account overlay	(agent, discovery_mode, filters, destinations, countries, account_id)	wholesale_feed_version_account, the wholesale signals feed payload when cache_scope: "account" was returned

Behavior.

• Requests without account always return cache_scope: "public". Callers cache under the public key.
• Requests with account return cache_scope: "public" OR "account" (agent MUST declare; no default).
  • "public": this account prices off the rate card. Caller MAY dedupe — the version and payload are the same as the unauthenticated view.
  • "account": this response carries account-specific overrides. Caller caches under the account overlay key.
• Agents MAY downgrade an account from "account" back to "public" — callers SHOULD interpret this as “this account no longer has overrides” and drop their overlay.

Conditional fetch with if_wholesale_feed_version. Send the token paired with whichever scope it was returned in. The agent compares against the current version for that scope. If the caller’s token belongs to an "account" scope but the agent responds with cache_scope: "public", that’s the downgrade signal.

Webhook invalidation. Wholesale feed webhook events declare applies_to.scope on *.priced and *.updated payloads. Agents MUST apply the same account/caller authorization predicate used by get_signals discovery_mode: "wholesale" when deciding which subscribers receive signal webhooks:

• applies_to: { scope: "public" } → invalidate the public-layer cache; all account overlays referencing that public version are also stale.
• applies_to: { scope: "account", account_ids: [...] } → invalidate only the named accounts’ overlays.
• applies_to: { scope: "account" } without account_ids → seller withholds the affected set; the per-subscriber scope filter routes the event only to subscribers whose principal is in the affected set.

See specs/wholesale-feed-webhooks.md §“Cache layering and event scoping” for the full webhook-side spec.

Incomplete array

When the agent cannot complete all work within the caller’s time_budget (or due to internal limits), the response includes incomplete — an array declaring what is missing. Callers can use estimated_wait to decide whether to retry with a larger budget.

Field	Type	Required	Description
scope	string	Yes	"signals": not all matching signals were returned. "pricing": signals returned but pricing is absent or unconfirmed. "wholesale_feed": in wholesale mode, full feed enumeration could not complete.
description	string	Yes	Human-readable explanation of what is missing and why.
estimated_wait	Duration	No	How much additional time would resolve this scope.

Iterative refinement

get_signals supports iterative refinement without a separate mode flag. The combination of signal_spec and signal_refs determines the operation:

Fields provided	Behavior
signal_spec only	Discovery — find signals matching the description
signal_refs only	Exact lookup — return specific signals by reference
signal_refs + signal_spec	Refinement — start from known signals, adjust per the spec
discovery_mode: "wholesale" (neither signal_spec nor signal_refs)	Wholesale — enumerate the agent’s full priced signals feed. See Wholesale signals feed.

To refine previous results, pass back the signal_ref values from signals you want to keep, and provide an updated signal_spec describing what to change:

{
  "$schema": "/schemas/signals/get-signals-request.json",
  "signal_spec": "Same audience but with broader coverage, ideally above 20%",
  "signal_refs": [
    {
      "scope": "data_provider",
      "data_provider_domain": "experian.com",
      "signal_id": "luxury_auto_intenders"
    }
  ],
  "destinations": [
    {
      "type": "agent",
      "agent_url": "https://wonderstruck.salesagents.com"
    }
  ],
  "countries": ["US"]
}

The signal agent uses the provided IDs as a starting point and the spec as adjustment guidance, returning signals that reflect both the original selection and the requested changes (e.g., broader segments from the same provider, or comparable segments from alternative providers with higher coverage).

Response - Multiple Signals Found

{
  "message": "I found 3 signals matching your luxury goods criteria. The best option is 'Affluent Shoppers' with 22% coverage, already live across all requested platforms. 'High Income Households' offers broader reach (35%) but requires activation on OpenX. All signals are priced between $2-4 CPM.",
  "context_id": "ctx-signals-abc123",
  "signals": [
    {
      "signal_ref": {
        "scope": "data_provider",
        "data_provider_domain": "acme-data.com",
        "signal_id": "affluent_shoppers"
      },
      "signal_agent_segment_id": "acme_affluent_shoppers",
      "name": "Affluent Shoppers",
      "description": "Users with demonstrated luxury purchase behavior",
      "signal_type": "marketplace",
      "data_provider": "Acme Data",
      "coverage_percentage": 22,
      "deployments": [
        {
          "type": "platform",
          "platform": "index-exchange",
          "account": "agency-123-ix",
          "is_live": true,
          "activation_key": {
            "type": "segment_id",
            "segment_id": "ix_agency123_acme_aff_shop"
          }
        },
        {
          "type": "platform",
          "platform": "openx",
          "account": "agency-123-ox",
          "is_live": true,
          "activation_key": {
            "type": "segment_id",
            "segment_id": "ox_agency123_affluent_789"
          }
        }
      ],
      "pricing_options": [
        {
          "pricing_option_id": "po_cpm_usd",
          "model": "cpm",
          "cpm": 3.50,
          "currency": "USD"
        }
      ]
    }
    // ... more signals
  ]
}

Response - Partial Success with Warnings

{
  "$schema": "/schemas/signals/get-signals-response.json",
  "status": "completed",
  "message": "Found 2 luxury signals, but encountered some platform limitations. The 'Premium Auto Shoppers' signal has limited reach due to data restrictions, and pricing data is unavailable for one platform. Review the warnings below for optimization suggestions.",
  "context_id": "ctx-signals-abc123",
  "cache_scope": "public",
  "wholesale_feed_version": "sig_v2026-05-18T08:00:00Z-public-rev12",
  "signals": [
    {
      "signal_ref": {
        "scope": "data_provider",
        "data_provider_domain": "experian.com",
        "signal_id": "premium_auto_shoppers"
      },
      "signal_agent_segment_id": "premium_auto_shoppers",
      "name": "Premium Auto Shoppers",
      "description": "High-value automotive purchase intenders",
      "signal_type": "marketplace",
      "data_provider": "Experian",
      "coverage_percentage": 8,
      "deployments": [
        {
          "type": "platform",
          "platform": "the-trade-desk",
          "is_live": true,
          "activation_key": {
            "type": "segment_id",
            "segment_id": "ttd_exp_auto_premium"
          }
        }
      ],
      "pricing_options": [
        {
          "pricing_option_id": "po_cpm_usd",
          "model": "cpm",
          "cpm": 4.50,
          "currency": "USD"
        }
      ]
    }
  ],
  "errors": [
    {
      "code": "PRICING_UNAVAILABLE",
      "message": "Pricing data temporarily unavailable for The Trade Desk platform",
      "field": "signals[0].pricing_options",
      "suggestion": "Retry in 15-30 minutes when platform pricing feed updates",
      "details": {
        "affected_platform": "the-trade-desk",
        "last_updated": "2025-01-15T12:00:00Z",
        "retry_after": 1800
      }
    },
    {
      "code": "PRICING_UNAVAILABLE", 
      "message": "Pricing data temporarily unavailable for Amazon DSP",
      "field": "filters.platforms",
      "suggestion": "Pricing will be available during activation, or try again later",
      "details": {
        "affected_platform": "amazon-dsp",
        "retry_after": 1800
      }
    }
  ]
}

Response - No Signals Found

{
  "$schema": "/schemas/signals/get-signals-response.json",
  "status": "completed",
  "message": "I couldn't find any signals matching 'underwater basket weavers' in the requested platforms. This appears to be a very niche audience. Consider broadening your criteria to 'craft enthusiasts' or 'hobby communities' for better results. Alternatively, we could create a custom signal for this specific audience.",
  "context_id": "ctx-signals-abc123",
  "cache_scope": "public",
  "wholesale_feed_version": "sig_v2026-05-18T08:00:00Z-public-rev12",
  "signals": []
}

Implementation Guide

Generating Signal Messages

The message field should provide actionable insights:

def generate_signals_message(signals, request):
    if not signals:
        return generate_no_signals_message(request.signal_spec)

    best_signal = find_best_signal(signals)

    if len(signals) == 1:
        signal = signals[0]
        deployment_status = get_deployment_summary(signal, request.destinations)
        pricing = signal.pricing_options[0] if signal.pricing_options else None
        if pricing:
            p = pricing.pricing
            if p.model == "cpm":
                price_commentary = f"Priced at ${p.cpm:.2f} CPM {p.currency}."
            elif p.model == "percent_of_media":
                cap = f", capped at ${p.max_cpm:.2f} CPM" if getattr(p, "max_cpm", None) else ""
                price_commentary = f"Priced at {p.percent}% of media spend{cap}."
            elif p.model == "flat_fee":
                price_commentary = f"Flat fee of {p.amount} {p.currency} per {p.period}."
            else:
                price_commentary = ""
        else:
            price_commentary = ""
        coverage_commentary = f" with {signal.coverage_percentage}% coverage" if getattr(signal, "coverage_percentage", None) is not None else ""
        return f"I found a perfect match: '{signal.name}' from {signal.data_provider}{coverage_commentary}. {deployment_status} {price_commentary}"
    else:
        return f"I found {len(signals)} signals matching your {extract_key_criteria(request.signal_spec)} criteria. {describe_best_option(best_signal)} {get_pricing_range(signals)}."

def get_deployment_summary(signal, requested_deployments):
    live = [d for d in signal.deployments if d.is_live]
    pending = [d for d in signal.deployments if not d.is_live]

    if not pending:
        return "Already live on all requested deployments, ready to use immediately."
    elif live:
        activation_time = max((d.estimated_activation_duration_minutes or 0) for d in pending)
        return f"Live on {len(live)} deployment(s). Activation on {len(pending)} more would take about {activation_time} minutes."
    else:
        return "Requires activation on all deployments, which typically takes 1-2 hours."