> ## Documentation Index > Fetch the complete documentation index at: https://docs.corti.ai/llms.txt > Use this file to discover all available pages before exploring further. # Export the observability trace for a context > Returns the execution traces for the context — LLM calls, tool executions, and token usage — in OpenInference format. Traces are ordered newest-first and paginated; each page returns up to `pageSize` traces with their spans inlined. ## OpenAPI ````yaml /agentic/auto-generated-openapi-v2.yml get /v2/agentic/contexts/{contextId}/trace openapi: 3.1.0 info: title: Corti Agent API v2 version: 2.0.0 summary: Manage agents and converse with them over the A2A v1.0 protocol. description: > Version 2 of the Corti Agents REST API. ## What's new in v2 - **Unified `connectors`** — a single, flat, discriminated array replaces the v1 split between `experts`, `mcpServers`, and sub-agents. The envelope is extensible: new connector kinds (`a2a`, `openapi`, `custom`) slot in behind the `type` discriminator without breaking changes. - **Clean CRUD verbs** — `POST` creates, `GET` fetches/lists, `PATCH` performs a *true* partial update, `DELETE` removes. v1's PATCH was effectively a PUT. - **First-class metadata** — `visibility`, `model`, `lifecycle` (now a body field, not a query param), and free-form `labels`. ## A2A v1.0 only The conversational surface speaks **A2A protocol version `1.0` exclusively** (both the `JSONRPC` and `HTTP+JSON` bindings). The deprecated v0.3 binding from v1 is intentionally not carried forward. Per A2A §3.6, clients MUST send the `A2A-Version` header (`Major.Minor`, e.g. `1.0`) on every request to the `/a2a/*` surface, or supply it as the `A2A-Version` query parameter. This surface implements `1.0` only; an absent header is treated as `1.0`. Patch versions MUST NOT be sent. The server echoes the negotiated version in the `A2A-Version` response header. ## Partial updates (PATCH) `PATCH` uses **JSON Merge Patch** semantics (RFC 7386), served under the `application/merge-patch+json` media type: - **Omit a field** → left unchanged. - **`null`** → cleared / reset to its default. - **Arrays** (e.g. `connectors`) → *replaced wholesale*, never merged. To mutate a single connector without replacing the whole array, use the `/connectors` sub-resource endpoints. ## Streaming (Server-Sent Events) The streaming A2A surfaces — `message/stream`, `tasks/subscribe`, and the streaming JSON-RPC methods — respond with `text/event-stream`. The `text/event-stream` media type carries a sequence of Server-Sent Events; the attached schema describes a single event frame. Every event's `data` field is a string carrying a JSON document (`contentMediaType: application/json`); `contentSchema` declares the parsed payload (`StreamResponse` for the `HTTP+JSON` binding, `JSONRPCResponse` for the `JSONRPC` binding). The optional `id` field carries the SSE event id used for resumption via the `Last-Event-ID` request header. contact: name: Corti API Support url: https://corti.ai email: support@corti.ai license: name: Proprietary url: https://corti.ai/terms termsOfService: https://corti.ai/terms servers: - url: https://api.{env}.corti.app description: Corti regional API gateway variables: env: default: eu enum: - eu - us description: Deployment region. security: - bearerAuth: [] tenantHeader: [] tags: - name: Agents description: Create, read, update, and delete agents. - name: Agent Card description: A2A-compliant agent discovery cards. - name: A2A description: Converse with an agent over the A2A v1.0 protocol. - name: Connectors description: Manage the connectors attached to an agent. - name: Usage description: Bucketed invocation history for an agent. - name: Contexts description: Conversational contexts and their tasks. - name: Registry description: Discoverable, pre-built connectors offered by the platform. - name: Artifacts description: Retrieve artifacts produced by tasks. - name: Feedback description: Collect human or automated feedback on tasks and messages. paths: /v2/agentic/contexts/{contextId}/trace: parameters: - $ref: '#/components/parameters/ContextId' get: tags: - Contexts summary: Export the observability trace for a context description: | Returns the execution traces for the context — LLM calls, tool executions, and token usage — in OpenInference format. Traces are ordered newest-first and paginated; each page returns up to `pageSize` traces with their spans inlined. operationId: ContextsGetTrace parameters: - $ref: '#/components/parameters/PageSize' - $ref: '#/components/parameters/PageToken' responses: '200': description: OpenInference trace JSON (paginated). content: application/json: schema: $ref: '#/components/schemas/ContextsTraceResponse' example: traces: - trace: id: 0192f4c85f3a7e8ab1c23d4e5f6a7b8c name: invoke_agent start_time: '2026-05-19T12:00:00Z' thread_id: ctx.0192f4c8-3d6b-7c4f-a02b-4d9e7f3c8b51 spans: - name: invoke_llm span_id: span.0192f4c8-6e1a-7f2b-9c3d-4e5f6a7b8c9d start_time: '2026-05-19T12:00:00Z' '400': $ref: '#/components/responses/BadRequest' '401': $ref: '#/components/responses/Unauthorized' '404': $ref: '#/components/responses/NotFound' components: parameters: ContextId: name: contextId in: path required: true description: Context identifier (prefixed UUIDv7). schema: $ref: '#/components/schemas/CommonContextIDValue' PageSize: name: pageSize in: query required: false description: Maximum number of items per page. schema: type: integer minimum: 1 maximum: 200 default: 50 PageToken: name: pageToken in: query required: false description: >- Opaque cursor from a prior response's `nextPageToken`. Omit on the first request. schema: type: string schemas: ContextsTraceResponse: type: object description: | A page of traces for a context in OpenInference format. Traces are ordered newest-first. required: - traces properties: traces: type: array items: $ref: '#/components/schemas/ContextsTraceItem' description: Traces for the context, newest first. examples: - - trace: id: 0192f4c85f3a7e8ab1c23d4e5f6a7b8c name: invoke_agent start_time: '2026-05-19T12:00:00Z' thread_id: ctx.0192f4c8-3d6b-7c4f-a02b-4d9e7f3c8b51 spans: - name: invoke_llm span_id: span.0192f4c8-6e1a-7f2b-9c3d-4e5f6a7b8c9d start_time: '2026-05-19T12:00:00Z' nextPageToken: $ref: '#/components/schemas/CommonNextPageToken' totalSize: $ref: '#/components/schemas/CommonTotalSize' CommonContextIDValue: type: string format: context-id description: >- Context identifier. Accepts `ctx.` or a bare UUIDv7 on input; always returned prefixed. pattern: ^(ctx\.)?[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}$ examples: - ctx.0192f4c8-3d6b-7c4f-a02b-4d9e7f3c8b51 ContextsTraceItem: type: object description: | A single trace with its inlined OpenInference spans. required: - trace - spans properties: trace: type: object description: The trace-level record. required: - id - name - start_time - thread_id properties: id: type: string description: Trace identifier (OTel trace ID — 32-char hex). name: type: string description: Human-readable trace name. start_time: type: string format: date-time description: When the trace started. end_time: type: - string - 'null' format: date-time description: When the trace ended; `null` if still in progress. input: type: object additionalProperties: true description: Trace-level input payload. output: type: object additionalProperties: true description: Trace-level output payload. metadata: type: object additionalProperties: true description: Free-form trace metadata. tags: type: array items: type: string description: Trace tags. thread_id: type: string description: Thread/context identifier. spans: type: array description: Spans in this trace, ordered by start time. items: $ref: '#/components/schemas/ContextsOpenInferenceSpan' CommonNextPageToken: type: - string - 'null' description: >- Opaque cursor to request the next page, or `null` if there are no more pages. CommonTotalSize: type: integer description: >- Total number of items matching the query, when known. Not currently populated by the server; treat as absent. example: 42 CommonErrorResponse: type: object description: > Corti management-plane error envelope, used by all non-A2A endpoints. - **Standard** — when the error chain contains at least one `PublicError`, `code` and `message` come from the outermost `PublicError` and `details` is merged across the whole chain (outer values take precedence). - **Fallback** — when the chain contains no `PublicError`, the response is a generic `500` carrying a `requestId` for support reference. - **Validation** — a single `PublicError` whose `details.validationErrors` lists the offending fields. Field names use camelCase on the wire (e.g. `requestId`, `howToFix`). The free-form `details` object may carry arbitrary caller-defined keys. Rate limiting (HTTP 429) is not yet implemented; the server does not emit a 429 response. required: - error properties: error: type: object required: - code - message properties: code: type: string description: Stable, machine-readable, SCREAMING_SNAKE_CASE error code. examples: - ASSIGNMENT_CONFLICT - VALIDATION_FAILED - INTERNAL_ERROR message: type: string description: Human-readable explanation. howToFix: type: string description: Optional guidance for the caller to resolve the error. details: type: object description: | Structured context, merged from every `PublicError` in the chain (outer values win). Omitted on the generic fallback response. additionalProperties: true properties: validationErrors: type: array description: Present when `code` is `VALIDATION_FAILED`. items: type: object required: - field - reason properties: field: type: string description: The field that failed validation. reason: type: string description: Why the field failed validation. requestId: type: string description: > Correlation ID from request middleware. Included only on the generic `500` fallback so consumers can quote it in support requests. description: The error object with code, message, and optional details. examples: - error: code: ASSIGNMENT_CONFLICT message: could not complete assignment details: assignment_id: asg_99 expert_id: exp_42 - error: code: VALIDATION_FAILED message: validation failed details: validationErrors: - field: email reason: invalid format - field: name reason: required - error: code: INTERNAL_ERROR message: internal server error requestId: req_abc123 ContextsOpenInferenceSpan: type: object description: A single span in an OpenInference trace. required: - name - span_id - start_time properties: name: type: string description: Human-readable span name. span_id: type: string description: Unique span identifier. parent_span_id: type: string description: Parent span id, omitted for the root span. start_time: type: string format: date-time description: When the span started. end_time: type: - string - 'null' format: date-time description: When the span ended; `null` if still in progress. attributes: type: object additionalProperties: true description: >- OpenInference span attributes. Key names and structure follow the OpenInference semantic conventions. examples: - llm.token_count.total: 42 - tool.name: icd10-lookup - input.value: 'chief complaint: chest pain' responses: BadRequest: description: The request was malformed. content: application/json: schema: $ref: '#/components/schemas/CommonErrorResponse' Unauthorized: description: Missing or invalid credentials. content: application/json: schema: $ref: '#/components/schemas/CommonErrorResponse' NotFound: description: The resource does not exist or is not visible to the caller. content: application/json: schema: $ref: '#/components/schemas/CommonErrorResponse' securitySchemes: bearerAuth: type: http scheme: bearer bearerFormat: JWT description: OAuth 2.0 / OIDC bearer token. tenantHeader: type: apiKey in: header name: Tenant-Name description: The tenant the request operates within. ````