Agentic Keychain Resolve API

openapi: 3.2.0
info:
  description: 'Agentic Keychain: Autonomous Trading Agents, Signed Results. This API is its skill registry (Agent Skill ROI Registry & Benchmarking); Keychain Trading''s results are read-only static files at https://agentickeychain.com/trading/results.json. This is a commercial registry. It sells some of the skills listed here, through Polar as merchant of record, and its operator earns money when one is bought. The evaluate answer can be "do not buy": it says so when the measured evidence does not cover your task, when the price is not recovered, or when a free alternative is enough. Every purchase ends in a checkout that a human completes. The same operations are MCP tools at /mcp. Fields this API version does not define can appear and have no meaning for a client that does not know them. A kind, delivery or pricing.model value that a client does not know means the capability is not supported by that client; it is never purchasable by inference.'
  title: Agentic Keychain Resolve API
  version: 0.2.0
servers:
- url: https://agentickeychain.com
tags:
- name: Resolve
paths:
  /v1/resolve:
    post:
      description: 'Describe a task in plain words and get one compact answer: whether a registry capability is worth using for it, from registry-run benchmarks only, and the next step. Use when: you have a task and no capability name, and want one decision word (use, do_not_use, use_free_alternative, insufficient_evidence, blocked_by_constraints or no_match) with its reason codes. Not for: the full arithmetic, the offer and the operator message (evaluate_capability with the same inputs), a plain list of matches (search_capabilities) or a capability you can already name (get_capability). Input: task, plus optional model, expected_runs, baseline_cost_estimate_usd, context and constraints; any other field is refused with the allowed list. Output: ak.resolve.v1 with decision, capability id, reason codes, the measured saving per run with its interval or null, the estimate at your own baseline with its break-even runs or null, confidence, evidence label, price, free alternatives, next action, evaluation_id and the tool that gives the detail. do_not_use and no_match are normal answers.'
      operationId: resolve_task
      requestBody:
        content:
          application/json:
            schema:
              additionalProperties: false
              properties:
                baseline_cost_estimate_usd:
                  description: What one run of this task costs you today without the capability, in USD.
                  exclusiveMinimum: 0
                  type: number
                constraints:
                  additionalProperties: false
                  description: Limits you set for the answer; a capability outside them is reported as blocked, with the reason code.
                  properties:
                    max_cost_usd:
                      description: The highest price in USD you accept; a capability priced above it is blocked (over_max_cost).
                      minimum: 0
                      type: number
                    minimum_confidence:
                      description: The lowest confidence in the measured saving you accept, from 0.5 to 0.999 (default 0.9); below it the answer is blocked (below_minimum_confidence).
                      maximum: 0.999
                      minimum: 0.5
                      type: number
                  type: object
                context:
                  additionalProperties:
                    type:
                    - string
                    - number
                    - integer
                    - boolean
                  description: Facts about your situation, keyed by the capability's parameter names (e.g. log_lines).
                  type: object
                expected_runs:
                  description: How many times you expect to perform this kind of task.
                  minimum: 1
                  type: integer
                model:
                  description: The model you run on, e.g. claude-sonnet-5. Savings are reported only for measured models.
                  type: string
                task:
                  description: Plain description of the task you are about to do (at most 4000 characters).
                  type: string
              required:
              - task
              type: object
        required: true
      responses:
        '200':
          content:
            application/json:
              schema:
                type: object
          description: Success
        default:
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Error'
          description: Error (ak.error.v1)
      summary: Is there a capability worth using for this task?
      tags:
      - Resolve
components:
  schemas:
    Error:
      properties:
        error:
          properties:
            code:
              type: string
            message:
              type: string
            retryable:
              type: boolean
          required:
          - code
          - message
          - retryable
          type: object
        schema:
          const: ak.error.v1
      required:
      - schema
      - error
      type: object