Documentation

One endpoint for chat, reasoning, image / video / 3D generation, and every registered skill — biometrics, emotion AI, vision, speech, and deep search. Pick a model, send a prompt, get a result.

Copy$ curl -X POST "https://model.mimicx.ai/api/v1/agent" \
  -H "Authorization: Bearer mx_live_..." \
  -H "Content-Type: application/json" \
  -d '{"model":"Nova","prompt":"generate a red tomato"}'

Introduction

The mimicx Model API is a single, unified endpoint. Instead of learning a different route for every capability, you send a prompt to /api/v1/agent and select a model. The model decides what to do:

• Darwin — a chat & reasoning LLM. Text conversation, planning, and text-based skills. Never generates media.
• Nova — everything Darwin does, plus automatic skill use: image, video & 3D generation, biometrics, emotion AI, vision, speech, deep search, and any skill your organization registers.

All traffic goes through https://model.mimicx.ai and is authenticated with a Bearer API key. Responses are JSON; add "stream": true for Server-Sent Events.

Authentication

Every request needs an API key in the Authorization header. Create and manage keys from the dashboard.

CopyAuthorization: Bearer mx_live_your_api_key

Models

List the live model catalog any time:

Copy$ curl "https://model.mimicx.ai/api/v1" \
  -H "Authorization: Bearer mx_live_..."

Chat & agent POST

POST /api/v1/agent — the primary endpoint. One prompt in, one combined result out. The response type tells you what came back.

Copy$ curl -X POST "https://model.mimicx.ai/api/v1/agent" \
  -H "Authorization: Bearer mx_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "Darwin",
    "prompt": "Explain quantum tunnelling in two sentences."
  }'

Response:

Copy{
  "type": "text",
  "text": "Quantum tunnelling is...",
  "model": "Darwin"
}

Request body

Streaming SSE

Set "stream": true to receive Server-Sent Events. Text arrives as delta events; the stream ends with [DONE].

Copy$ curl -N -X POST "https://model.mimicx.ai/api/v1/agent" \
  -H "Authorization: Bearer mx_live_..." \
  -H "Content-Type: application/json" \
  -d '{"model":"Darwin","prompt":"Write a haiku","stream":true}'

data: {"type":"delta","text":"Silent"}
data: {"type":"delta","text":" code..."}
data: [DONE]

Media and skill results can't stream token-by-token — they arrive as a single final event (an image, a job reference, etc.) before [DONE].

Images, video & 3D Nova

With model: "Nova", a generation request returns base64 media.

Copy$ curl -X POST "https://model.mimicx.ai/api/v1/agent" \
  -H "Authorization: Bearer mx_live_..." \
  -H "Content-Type: application/json" \
  -d '{"model":"Nova","prompt":"a watercolor fox, side profile"}'

Long-running jobs beta

Some skills — builds, large renders, training — can't finish inside a request. They return a job reference; subscribe to its stream for progress.

Copy{
  "type": "job",
  "skill": "firmware_build",
  "job": {
    "id": "bld_7f3a",
    "status": "queued",
    "stream": "/api/v1/firmware/jobs/bld_7f3a"
  }
}

The stream emits status, progress, diagnostic, and a final done (with the artifact) or error event.

Skills & tools

A skill is a callable capability. Skill-capable models (Nova) invoke them automatically from your prompt, and you can always call one directly by name.

List skills GET

Copy$ curl "https://model.mimicx.ai/api/v1/skills" \
  -H "Authorization: Bearer mx_live_..."

Invoke a skill POST

POST /api/v1/skills/{name}/invoke — call a skill directly, bypassing model routing. Pass inputs under args.

Copy$ curl -X POST "https://model.mimicx.ai/api/v1/skills/create_pdf/invoke" \
  -H "Authorization: Bearer mx_live_..." \
  -H "Content-Type: application/json" \
  -d '{"args":{"title":"Report","sections":[]}}'

Biometrics

Face, iris, liveness, gait, footprint, person re-identification and object signatures. Send an image with the request; the router selects the right skill, or target one directly.

Copy$ curl -X POST "https://model.mimicx.ai/api/v1/agent" \
  -H "Authorization: Bearer mx_live_..." \
  -H "Content-Type: application/json" \
  -d '{"prompt":"detect the face in this image","image_b64":"iVBOR..."}'

Enrollment & identification POST

Register people, then recognize them later. Enrollment stores a biometric template under a user_id; identification matches a new sample against everyone enrolled and returns the best match with a similarity score.

Enroll a face — call the skill directly for deterministic results. A user_id is required; multiple enroll calls for the same id refine the template (running average).

Copy$ curl -X POST "https://model.mimicx.ai/api/v1/skills/biometrix_enroll/invoke" \
  -H "Authorization: Bearer mx_live_..." \
  -H "Content-Type: application/json" \
  -d '{"args":{"action":"enroll","task":"face","user_id":"maria_02","image_b64":"iVBOR..."}}'

# → {"status":"ok","enrolled":"maria_02","task":"face","samples":1,"dim":512}

Identify a face — returns the matched user_id and a cosine similarity score (0–1). Pass an optional threshold (default 0.5); below it the result is no_match.

Copy$ curl -X POST "https://model.mimicx.ai/api/v1/skills/biometrix_identify/invoke" \
  -H "Authorization: Bearer mx_live_..." \
  -H "Content-Type: application/json" \
  -d '{"args":{"action":"identify","task":"face","image_b64":"iVBOR...","threshold":0.5}}'

# → {"status":"match","user_id":"maria_02","score":0.87,"threshold":0.5}

Two management actions are also available on biometrix_enroll: {"action":"list_enrolled"} returns enrolled ids, and {"action":"delete_enrolled","user_id":"..."} removes one.

Consent & privacy. Only enroll biometrics with the person's consent and a lawful basis. Templates are personal data and cannot be reset like a password — store and transmit them securely, and delete on request.

Emotion AI

Text sentiment, facial expression, voice tone, and multimodal fusion.

Endpoint reference

The full machine-readable spec is at /dev/openapi, with an interactive explorer at Swagger UI.

Errors

Errors return a JSON body with a detail message and a standard HTTP status.

Rate limits

Limits depend on your plan and are enforced per API key. A 429 means you've exceeded your current rate — retry with exponential backoff. See your plan & usage in the dashboard.

mimicx Model API · Built for developers · Dashboard · mimicx.ai