> ## Documentation Index > Fetch the complete documentation index at: https://docs.agentsfleet.net/llms.txt > Use this file to discover all available pages before exploring further. # Keep each tool call's full output > Keeps the full arguments and output of each finished tool call, so an operator can open everything a call took and returned. Post before the report. A body is at most 262144 bytes, and each record's arguments and output at most 65536 bytes each. An event keeps at most 1 MiB of records. A record that breaks a bound or does not fit is counted in `skipped_count` and the rest are kept. Posting a call again replaces its record, so a retry is safe. A lease the fleet has moved past is refused and keeps nothing. ## OpenAPI ````yaml https://raw.githubusercontent.com/agentsfleet/agentsfleet/main/public/openapi.json post /v1/runners/me/leases/{lease_id}/tool-calls openapi: 3.1.0 info: title: agentsfleet Control Plane API description: API for managing workspaces, fleets, triggers, and runs. contact: name: agentsfleet url: https://agentsfleet.net license: name: MIT identifier: MIT version: 1.0.0 servers: - url: https://api.agentsfleet.net description: Production security: [] paths: /v1/runners/me/leases/{lease_id}/tool-calls: post: tags: - Runners summary: Keep each tool call's full output description: >- Keeps the full arguments and output of each finished tool call, so an operator can open everything a call took and returned. Post before the report. A body is at most 262144 bytes, and each record's arguments and output at most 65536 bytes each. An event keeps at most 1 MiB of records. A record that breaks a bound or does not fit is counted in `skipped_count` and the rest are kept. Posting a call again replaces its record, so a retry is safe. A lease the fleet has moved past is refused and keeps nothing. operationId: runner_record_tool_calls parameters: - name: lease_id in: path description: '`UUIDv7` of the lease.' required: true schema: type: string requestBody: content: application/json: schema: $ref: '#/components/schemas/ToolCallRecordsRequest' required: true responses: '200': description: OK content: application/json: schema: $ref: '#/components/schemas/ToolCallRecordsStored' '400': description: The request could not be read content: application/problem+json: schema: $ref: '#/components/schemas/ProblemBody' '401': description: No credential, or one this route does not accept content: application/problem+json: schema: $ref: '#/components/schemas/ProblemBody' '403': description: The credential is good and lacks the capability this route requires content: application/problem+json: schema: $ref: '#/components/schemas/ProblemBody' '404': description: No such resource, or none this caller may see content: application/problem+json: schema: $ref: '#/components/schemas/ProblemBody' '409': description: The request conflicts with the resource's current state content: application/problem+json: schema: $ref: '#/components/schemas/ProblemBody' '413': description: The payload is over this route's ceiling content: application/problem+json: schema: $ref: '#/components/schemas/ProblemBody' '429': description: The instance is at its ceiling content: application/problem+json: schema: $ref: '#/components/schemas/ProblemBody' '500': description: The daemon failed to answer content: application/problem+json: schema: $ref: '#/components/schemas/ProblemBody' '503': description: A dependency this route needs is unreachable content: application/problem+json: schema: $ref: '#/components/schemas/ProblemBody' security: - RunnerBearerAuth: [] components: schemas: ToolCallRecordsRequest: type: object description: '`POST /v1/runners/me/leases/{lease_id}/tool-calls` request.' required: - fencing_token - calls properties: calls: type: array items: $ref: '#/components/schemas/ToolCallRecord' description: The finished calls to keep. A call posted again replaces its record. fencing_token: type: integer format: int64 description: >- The lease's fencing token; a holder the fleet has superseded is refused. minimum: 0 additionalProperties: false ToolCallRecordsStored: type: object description: What a post kept. required: - stored_count - skipped_count properties: skipped_count: type: integer description: >- Records refused for their shape, a bound, or the event's budget. Beta: this field may change shape before it is stable. minimum: 0 x-stability: beta stored_count: type: integer description: >- Records written or replaced. Beta: this field may change shape before it is stable. minimum: 0 x-stability: beta ProblemBody: type: object description: The body every refusal carries, under `application/problem+json`. required: - docs_uri - title - detail - error_code - request_id properties: action_id: type: - string - 'null' description: 'On that 409 only: the action the gate held.' current_state: type: - string - 'null' description: 'On a 409 only: the state that forbade the transition.' detail: type: string description: One sentence on what was refused and why. docs_uri: type: string description: Where the error code is documented. error_code: type: string description: The registry code, stable across releases. etag: type: - string - 'null' description: >- On a 412 only: the resource's current entity tag, to refetch and retry. gate_id: type: - string - 'null' description: 'On an answered approval gate''s 409 only: the gate that was answered.' outcome: type: - string - 'null' description: 'On that 409 only: the answer that stands.' request_id: type: string description: The request this refusal answers, for support. resolved_at: type: - integer - 'null' format: int64 description: 'On that 409 only: when the standing answer was given.' resolved_by: type: - string - 'null' description: 'On that 409 only: who gave it.' title: type: string description: A short name for the refusal, safe to show a person. user_message: type: - string - 'null' description: A curated sentence for end users, where the code has one. ToolCallRecord: type: object description: One finished call as the runner records it. required: - call_number - arguments - truncated_arguments - output - output_line_count - truncated properties: arguments: type: object description: |- Every argument the call was made with, secret values masked. At most 65536 bytes encoded, with no NUL character. call_number: type: integer format: int64 description: |- The call's number in its run, from 1: the `n` of the `{fence}:{n}` call id the live frames and the trace carry. maximum: 9223372036854776000 minimum: 1 output: type: string description: |- Everything the call returned, at most 65536 bytes, with no NUL character. output_line_count: type: integer format: int64 description: How many lines the whole output had, before any cut. minimum: 0 truncated: type: boolean description: Whether the runner cut the output to fit. truncated_arguments: type: boolean description: Whether the runner cut the arguments to fit. additionalProperties: false securitySchemes: RunnerBearerAuth: type: http scheme: bearer bearerFormat: agt_r description: The opaque agt_r token minted when the runner enrols (POST /v1/runners) ```` This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.