Mnemom Aletheia Support API

In-product Aletheia assistant MVP (issue #2195) — chat + voice adapters over one shared answer core. Gated behind ALETHEIA_SUPPORT_MVP_ENABLED + the ALETHEIA_SUPPORT_ALLOWLIST identity allowlist (whose default is the whole @mnemom.ai domain, and which may be set to "*" to admit every authenticated identity), fail-closed 404 for every non-allowed caller. Dark by default; whether it is customer-facing depends on that env var per deployment.

GitHub OpenAPI

Operations 3

POST /aletheia-support/chat Send a chat-native turn to the in-product Aletheia assistant #
POST /aletheia-support/voice Send a voice-native turn to the in-product Aletheia assistant #
GET /aletheia/gate Check whether the caller may see the in-product Aletheia assistant (DARK MVP) #

Work with this as data

Every API here is available over the APIs.io API and to AI agents over MCP.

MCP server

One button, every client — Claude, Cursor, VS Code and the rest.

https://apis.io/mcp

Tools for apis

7 MCP tools reach this
  • find_apisBrowse and filter every API in the catalog.
  • get_api_artifactsOne API's artifacts, grouped by type.
  • get_openapiThe primary OpenAPI for this API.
  • find_similar_apisAPIs that look like this one.
  • apis_io_searchSTART HERE — APIs, providers and tags for one query, each with its total.
  • resolveTurn a domain, URL or GitHub org into the provider it belongs to.
  • find_cohortsEvery scored population of providers in the catalog.
All 92 tools →

Call it yourself

curl for this page
This API
curl "https://apis.io/api/v1/apis/mnemom-ai:mnemom-ai-aletheia-support-api"
All apis
curl "https://apis.io/api/v1/apis?limit=25"

Calls without X-API-Key need no key. Calls with it need a paid plan — get a key.

Get an API key

Free tier, no form to fill in. Signing in shares your email address with us — we store it to create your key and to recognise you if you sign in with another provider. By signing up you are subscribing to the APIs.io and API Evangelist newsletters — you can unsubscribe from any issue. See our Privacy Policy and Terms.

A second provider on the same verified email joins the account you already have.

OpenAPI Specification

mnemom-ai-aletheia-support-api-openapi.yml Raw ↑
openapi: 3.2.0
info:
  title: Mnemom Aletheia Support API
  description: Trust infrastructure for AI agents. Transparent alignment verification, behavioral drift detection, and accountability primitives.
  version: 1.0.0
  contact:
    name: Mnemom
    url: https://mnemom.ai
    email: support@mnemom.ai
  license:
    name: Apache 2.0
    url: https://www.apache.org/licenses/LICENSE-2.0
servers:
- url: https://api.mnemom.ai/v1
  description: Production
security:
- BearerAuth: []
- ApiKeyAuth: []
tags:
- name: Aletheia Support
  description: 'In-product Aletheia assistant MVP (issue #2195) — chat + voice adapters over one shared answer core. Gated behind ALETHEIA_SUPPORT_MVP_ENABLED + the ALETHEIA_SUPPORT_ALLOWLIST identity allowlist (whose default is the whole @mnemom.ai domain, and which may be set to "*" to admit every authenticated identity), fail-closed 404 for every non-allowed caller. Dark by default; whether it is customer-facing depends on that env var per deployment.'
paths:
  /aletheia-support/chat:
    post:
      operationId: aletheiaSupportChat
      summary: Send a chat-native turn to the in-product Aletheia assistant
      description: 'DARK MVP (issue #2195): normalizes the chat-native `message` and runs it through the SAME answer-core the voice adapter uses, returning a chat-shaped result. Gated behind `ALETHEIA_SUPPORT_MVP_ENABLED` + the `ALETHEIA_SUPPORT_ALLOWLIST` identity allowlist (src/aletheia-support/gate.ts; defaults to the whole @mnemom.ai domain, and may be set to `*` to admit every authenticated identity) — every caller the gate doesn''t allow (no user, wrong identity, or flag off) gets an identical 404, never a 401/403 that would reveal the route''s existence or gating reason. Binds the real grounded answer core (`makeGroundedAnswerCore`); the fail-closed `denyAnswerCore` it originally shipped with was replaced when MNE-1937 landed.'
      tags:
      - Aletheia Support
      security: []
      requestBody:
        required: true
        content:
          application/json:
            schema:
              type: object
              required:
              - message
              properties:
                message:
                  type: string
                  description: The caller's chat turn. Must be a non-empty string.
                session_id:
                  type: string
                  description: Optional chat session id.
                locale:
                  type: string
                  description: Optional BCP-47 locale.
      responses:
        '200':
          description: The core's answer, chat-shaped. `refused:true` means the core (or the adapter's length guard) withheld an ungrounded answer; `answer` is then a fixed canned refusal string, never model-generated.
          content:
            application/json:
              schema:
                type: object
                required:
                - answer
                - grounding
                - refused
                properties:
                  answer:
                    type: string
                  grounding:
                    type: array
                    items:
                      type: object
                      required:
                      - id
                      properties:
                        id:
                          type: string
                        title:
                          type: string
                        url:
                          type: string
                        snippet:
                          type: string
                  refused:
                    type: boolean
                  refusalReason:
                    type: string
        '400':
          $ref: '#/components/responses/BadRequest'
        '403':
          $ref: '#/components/responses/CrossOriginRejected'
        '404':
          $ref: '#/components/responses/NotFound'
        '429':
          $ref: '#/components/responses/TooManyRequests'
        '500':
          $ref: '#/components/responses/InternalServerError'
  /aletheia-support/voice:
    post:
      operationId: aletheiaSupportVoice
      summary: Send a voice-native turn to the in-product Aletheia assistant
      description: 'DARK MVP (issue #2195): normalizes the voice-native `transcript` and runs it through the SAME answer-core the chat adapter uses, returning a voice-shaped (`speech`) result. Same gate as the chat sibling — see that operation''s description. Binds the real grounded answer core (`makeGroundedAnswerCore`); the fail-closed `denyAnswerCore` it originally shipped with was replaced when MNE-1937 landed.'
      tags:
      - Aletheia Support
      security: []
      requestBody:
        required: true
        content:
          application/json:
            schema:
              type: object
              required:
              - transcript
              properties:
                transcript:
                  type: string
                  description: The caller's spoken turn (ASR transcript). Must be a non-empty string.
                call_id:
                  type: string
                  description: Optional voice call id.
                locale:
                  type: string
                  description: Optional BCP-47 locale.
      responses:
        '200':
          description: The core's answer, voice-shaped. `refused:true` means the core (or the adapter's length guard) withheld an ungrounded answer; `speech` is then a fixed canned refusal string, never model-generated.
          content:
            application/json:
              schema:
                type: object
                required:
                - speech
                - grounding
                - refused
                properties:
                  speech:
                    type: string
                  grounding:
                    type: array
                    items:
                      type: object
                      required:
                      - id
                      properties:
                        id:
                          type: string
                        title:
                          type: string
                        url:
                          type: string
                        snippet:
                          type: string
                  refused:
                    type: boolean
                  refusalReason:
                    type: string
        '400':
          $ref: '#/components/responses/BadRequest'
        '403':
          $ref: '#/components/responses/CrossOriginRejected'
        '404':
          $ref: '#/components/responses/NotFound'
        '429':
          $ref: '#/components/responses/TooManyRequests'
        '500':
          $ref: '#/components/responses/InternalServerError'
  /aletheia/gate:
    get:
      operationId: aletheiaGate
      summary: Check whether the caller may see the in-product Aletheia assistant (DARK MVP)
      description: The single check the front-of-house widget polls BEFORE rendering (MNE-1930/1931) — the backend is the ONLY authority; the client never inspects the flag or allowlist itself. Reuses the exact `supportAccessAllowed` predicate `POST /v1/aletheia-support/{chat,voice}` enforce, so this can never drift out of sync with what those routes would actually allow. Unlike the chat/voice routes, ALWAYS returns 200 — the gate-check's own existence isn't sensitive (it reveals nothing about grounding/answers) — with `allowed` reflecting the real decision (`false` for flag-off, non-allowlisted, or no caller at all).
      tags:
      - Aletheia Support
      security: []
      responses:
        '200':
          description: Always returned, for every caller.
          content:
            application/json:
              schema:
                type: object
                required:
                - allowed
                properties:
                  allowed:
                    type: boolean
        '429':
          $ref: '#/components/responses/TooManyRequests'
        '500':
          $ref: '#/components/responses/InternalServerError'
components:
  responses:
    InternalServerError:
      description: Server error — request was well-formed but the server failed to fulfill it. Typically a downstream dependency (DB / RPC / external API) returned an unexpected error.
      content:
        application/json:
          schema:
            $ref: '#/components/schemas/Error'
    TooManyRequests:
      description: Rate-limit exceeded. The global per-IP limiter (100 requests/minute, applied to every `/v1/*` route) rejected this request. Back off until the window resets — `Retry-After` carries the cooldown in seconds and `X-RateLimit-Reset` the absolute reset time.
      headers:
        Retry-After:
          description: Seconds to wait before retrying.
          schema:
            type: integer
            minimum: 1
        X-RateLimit-Limit:
          description: Requests permitted per window.
          schema:
            type: integer
        X-RateLimit-Remaining:
          description: Requests remaining in the current window (0 on a 429).
          schema:
            type: integer
            minimum: 0
        X-RateLimit-Reset:
          description: Unix epoch seconds at which the current window resets.
          schema:
            type: integer
      content:
        application/json:
          schema:
            $ref: '#/components/schemas/Error'
    CrossOriginRejected:
      description: Cross-origin write refused (`cross_origin_rejected`). A state-changing request that carries the browser session cookie must come from a first-party origin (`Origin`, or failing that `Referer`) and must not send a form-style body (text/plain, urlencoded, multipart) the operation does not declare. Bearer and API-key callers are never subject to this check.
      content:
        application/json:
          schema:
            $ref: '#/components/schemas/Error'
    BadRequest:
      description: Bad request
      content:
        application/json:
          schema:
            $ref: '#/components/schemas/Error'
    NotFound:
      description: Resource not found
      content:
        application/json:
          schema:
            $ref: '#/components/schemas/Error'
  schemas:
    Error:
      type: object
      description: Canonical error envelope (ADR-API-001 conv 1). `error` is always an object — never a bare string. Every 4xx/5xx response across the API conforms to this shape; the runtime helper is `src/http-errors.ts::buildErrorBody`.
      required:
      - error
      properties:
        error:
          type: object
          required:
          - code
          - message
          properties:
            code:
              type: string
              pattern: ^[a-z][a-z0-9_]*$
              description: 'Stable, machine-matchable failure identifier (lowercase snake_case). Clients may branch on this; the string is part of the contract.


                **Status-class defaults** — emitted when no caller code is supplied (`errorCodeForStatus(status)`): `bad_request` (400), `unauthorized` (401), `forbidden` (403), `not_found` (404), `method_not_allowed` (405), `conflict` (409), `gone` (410), `precondition_failed` (412), `payload_too_large` (413), `unsupported_media_type` (415), `unprocessable_entity` (422), `precondition_required` (428), `rate_limited` (429), `internal_error` (500), `not_implemented` (501), `bad_gateway` (502), `service_unavailable` (503), `gateway_timeout` (504). Fallback `error` for unmapped statuses.


                **Caller-supplied codes** — handlers may pass an explicit `code` for a specific failure class. Examples: `agent_not_found`, `invalid_hash_proof`, `already_linked`, `idempotency_conflict`, `feature_gated`, `schema_validation_failed`, `no_token`, `bad_canonical_payload`.


                **Care-framed sub-resource codes** — the cards-as-primitive surface passes its stable care code-string straight through as `error.code`. Examples: `if_match_absent`, `if_match_stale`, `if_match_malformed`, `primitive_validation_failed`.'
            message:
              type: string
              description: Human-readable, care-framed explanation of the failure.
            details:
              description: 'Optional structured context for the failure (any JSON value: object, array, or primitive). Common shapes: validation findings list, idempotency-conflict diff, `{presented_etag, current_etag}` on a stale `If-Match`, etc. Mirrors the helper''s `details?: unknown`.'
  securitySchemes:
    BearerAuth:
      type: http
      scheme: bearer
      bearerFormat: JWT
      description: 'Supabase JWT token in Authorization: Bearer header'
    ApiKeyAuth:
      type: apiKey
      in: header
      name: X-Mnemom-Api-Key
      description: Mnemom API key (mnm_... format)
    ServiceRole:
      type: http
      scheme: bearer
      bearerFormat: JWT
      description: Supabase service role bearer JWT. Used by /blog admin operations that talk directly to Supabase. NOT used by /internal/* or /v1/arena/internal/* — those use the custom-header schemes below.
    InternalKeyAuth:
      type: apiKey
      in: header
      name: X-Internal-Key
      description: Internal service-to-service shared secret, timing-safe compared against INTERNAL_API_KEY. Used by /internal/* gateway routes (compose, agents, webhooks emit).
    ServiceKeyAuth:
      type: apiKey
      in: header
      name: X-Service-Key
      description: Service-to-service shared secret, timing-safe compared against MNEMOM_SERVICE_KEY. Used by /v1/arena/internal/*, /v1/on-chain/*, and the /v1/internal/* readers consumed by the reputation cron.
    CoherenceTokenAuth:
      type: http
      scheme: bearer
      description: Per-cell coherence bearer token. Provisioned in 1Password vault `mnemom-coherence` (one item per receiver). Value compared against COHERENCE_<NAME>_SERVICE_TOKEN_EXPECTED. Used by /v1/internal/coherence/* and /v1/internal/governance/dispatch.
    ArenaCandidateTokenAuth:
      type: http
      scheme: bearer
      description: Dedicated arena writer token (ADR-004 §"Arena dedicated writer token"). Compared against ARENA_RECIPE_CANDIDATE_TOKEN. The token's only capability is writing recipe candidates with writer_identity='arena-bypass' (stamped server-side). Used by POST /v1/internal/recipe-candidates.
    CustomerVoiceTokenAuth:
      type: http
      scheme: bearer
      description: Customer Voice API bearer token. Timing-safe compared against CUSTOMER_VOICE_API_TOKEN. Grants access to all /v1/internal/feedback/* endpoints (MNE-1876). No customer-facing or reporter-read capability.
    CookieAuth:
      type: apiKey
      in: cookie
      name: mnemom_session
      description: 'HttpOnly, Secure, SameSite=Lax cookie issued by /v1/auth/sign-in (or the SSO / email-callback flows). The value is an AES-256-GCM-encrypted blob of {access_token, refresh_token, issued_at, auth_method}. Browser clients include this automatically with `credentials: "include"`.'
    LicenseJwtAuth:
      type: http
      scheme: bearer
      bearerFormat: JWT
      description: License JWT (HS256, HMAC-signed with LICENSE_SIGNING_SECRET). Used by self-hosted deployment instances to authenticate `/v1/deployments/heartbeat`. Not interchangeable with the Supabase user JWT.
    AgentAuth:
      type: apiKey
      in: header
      name: X-Mnemom-Agent-Proof
      description: 'Agent possession proof: the full SHA-256 hex digest of `${apiKey}|${agentName}` (or `${apiKey}` for an unnamed singleton agent) — the same `hash_proof` primitive used by POST /v1/agents claim, link-agent, and verify-binding. The server resolves the agent by `agent_hash = proof[:16]`. The agent''s *own* identity (ADR-API-001 third auth class) — distinct from ApiKeyAuth (`X-Mnemom-Api-Key`, which resolves a user/account, not an agent).'