Build on the complete ELTEX APIs.

01 · QUICKSTART

Connect with an Agent Plan key.

Use a server-side API key or an OAuth access token. Never expose a secret API key in browser code. Every /v1 response includes X-Request-ID; include it in support reports.

Base URL: https://eltexlabs.com/v1
Authorization: Bearer sk-eltex-...
Content-Type: application/json

curl https://eltexlabs.com/v1/chat/completions \
  -H "Authorization: Bearer $ELTEX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"eltex/smart","messages":[{"role":"user","content":"Explain RAG simply."}]}'

Download Agent Plan reference Download Agent State reference Download RunGuard reference Download Agentic Wallet reference Download agent skill

02 · MODEL DISCOVERY

List models available to this account.

GET /v1/models

{
  "object": "list",
  "data": [
    { "id": "eltex/smart", "object": "model", "owned_by": "eltex" },
    { "id": "MODEL_ID_FROM_GET_MODELS", "object": "model", "owned_by": "eltex" }
  ]
}

The response is filtered by the authenticated account and active plan. Treat it as the source of truth for model IDs. The catalog reports availability; it does not yet expose a per-model capability matrix.

03 · CHAT COMPLETIONS

Request and response schema.

POST /v1/chat/completions

The non-stream response preserves the OpenAI-compatible completion and adds an eltex object:

{
  "id": "chatcmpl_...",
  "object": "chat.completion",
  "model": "selected-provider-model",
  "choices": [{
    "index": 0,
    "message": { "role": "assistant", "content": "Hello!" },
    "finish_reason": "stop"
  }],
  "usage": {
    "prompt_tokens": 12,
    "completion_tokens": 4,
    "total_tokens": 16
  },
  "eltex": {
    "classification": "general_qa",
    "complexity": "simple",
    "requested_model": "eltex/smart",
    "selected_model": "MODEL_ID_FROM_GET_MODELS",
    "routing_mode": "smart",
    "routing_decision": "preclassifier_smart_route",
    "preclassifier_skipped": false,
    "credits_charged": 1.25,
    "input_tokens": 12,
    "output_tokens": 4,
    "estimated_usage": false,
    "request_id": "chatcmpl_..."
  }
}

Classification and routing fields are operational metadata, not security boundaries. credits_charged is the Agent Plan credit charged for the completion.

04 · PARAMETER COMPATIBILITY

Tools, structured output, and multimodal input.

ELTEX passes OpenAI-compatible generation parameters to the selected route. Capability support remains model-dependent; test the exact model and payload before production use.

Tool calling

{
  "model": "MODEL_ID_FROM_GET_MODELS",
  "messages": [{ "role": "user", "content": "Weather in Bangkok?" }],
  "tools": [{
    "type": "function",
    "function": {
      "name": "get_weather",
      "parameters": {
        "type": "object",
        "properties": { "city": { "type": "string" } },
        "required": ["city"],
        "additionalProperties": false
      }
    }
  }],
  "tool_choice": "auto"
}

Structured output

{
  "model": "MODEL_ID_FROM_GET_MODELS",
  "messages": [{ "role": "user", "content": "Return a project risk." }],
  "response_format": {
    "type": "json_schema",
    "json_schema": {
      "name": "risk",
      "strict": true,
      "schema": {
        "type": "object",
        "properties": { "level": { "type": "string" }, "summary": { "type": "string" } },
        "required": ["level", "summary"],
        "additionalProperties": false
      }
    }
  }
}

Multimodal input

{
  "model": "eltex/smart",
  "messages": [{
    "role": "user",
    "content": [
      { "type": "text", "text": "Describe this image." },
      { "type": "image_url", "image_url": { "url": "https://example.com/image.png" } }
    ]
  }]
}

The selected model must support the requested tools, schema mode, or media. Use the separate Elixir API to generate images and videos.

05 · ROUTING

Smart, Preferred, and Force.

Markers are inspected only in the latest user message. ELTEX removes the first $*, $?, or $! it finds. There is no backslash escape syntax. To show a literal marker in that message, write it non-contiguously, such as $ !, or place the exact literal in an earlier system/developer message.

06 · STREAMING

SSE chunks, usage, and failures.

Set stream: true. ELTEX returns text/event-stream with OpenAI-compatible chunks and a terminal [DONE].

data: {"id":"chatcmpl_...","object":"chat.completion.chunk","choices":[{"index":0,"delta":{"content":"Hello"},"finish_reason":null}]}

data: {"id":"chatcmpl_...","object":"chat.completion.chunk","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

data: [DONE]

Request "stream_options":{"include_usage":true} when the selected model supports a usage chunk. ELTEX records final usage after the upstream stream is consumed successfully. The non-stream eltex metadata object is not added to stream chunks.

Errors before streaming starts use the JSON error schema. Mid-stream failures may arrive as an SSE error event or an early connection close. There is no resume cursor; treat an early close as indeterminate and reconcile usage before retrying.

07 · AGENT CONTROL

Usage, capacity, Burst, and Reset.

Usage

GET /v1/usage?range=week

{
  "object": "agent_plan.usage",
  "range": "week",
  "requests": 42,
  "input_tokens": 12000,
  "output_tokens": 3400,
  "cost_usd": 0.124,
  "cost_idr": 2046,
  "internal_units": 1.24,
  "credits_used": 124,
  "burst_credits_used": 0,
  "reset_credits_used": 20
}

range accepts day, week, month, or all. Omitted or unrecognized values currently fall back to all.

Capacity

GET /v1/capacity

{
  "object": "agent_plan.capacity",
  "plan": "Plus",
  "period": {
    "remaining_percent": 72.5,
    "used_percent": 27.5,
    "remaining_credits": 3262.5,
    "limit_credits": 4500,
    "resets_at": "2026-07-29T20:00:00.000Z"
  },
  "weekly": {
    "remaining_percent": 81.2,
    "used_percent": 18.8,
    "resets_at": "2026-08-03T00:00:00.000Z",
    "timezone": "Asia/Bangkok"
  },
  "burst": { "active_credits": 500, "total_credits": 500, "auto_active": true },
  "reset": { "stored_count": 2, "usable_count": 1, "active": false, "active_period_credits": 0 }
}

Burst status

GET /v1/burst

{
  "object": "agent_plan.burst",
  "auto_active": true,
  "active_credits": 500,
  "total_credits": 500,
  "used_after": ["regular plan capacity", "active Reset Ticket capacity"]
}

Reset inventory

GET /v1/reset

{
  "object": "agent_plan.reset",
  "plan": "Plus",
  "stored_count": 2,
  "usable_count": 1,
  "active": false,
  "active_period_credits": 0,
  "tickets": [{
    "label": "Plus Reset Ticket",
    "plan_key": "plus",
    "source": "purchase",
    "usable_now": true,
    "stored_at": "2026-07-29T10:00:00.000Z"
  }]
}

Activate one stored Reset Ticket

curl -X POST https://eltexlabs.com/v1/reset/use \
  -H "Authorization: Bearer $ELTEX_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: reset-job-<stable-job-id>" \
  -d '{"reason":"capacity exhausted during an active user-approved job"}'

{
  "object": "agent_plan.reset_activation",
  "activated": true,
  "ticket": "Plus Reset Ticket",
  "plan": "Plus",
  "period_remaining_percent": 100,
  "weekly_remaining_percent": 100,
  "period_resets_at": "2026-07-29T20:00:00.000Z",
  "weekly_resets_at": "2026-08-03T00:00:00.000Z",
  "idempotent_replay": false
}

08 · AGENT STATE API

Keep agent work consistent across sessions.

Agent State stores agents, task state, event history, checkpoints, resume runs, private managed files in Cloudflare R2, and references to external artifacts. It is available on every paid Agent Plan; Free-plan credentials return 403 agent_runtime_paid_plan_required.

POST   /v1/agents
GET    /v1/agents?status=active&limit=25&cursor=OPAQUE_CURSOR
GET    /v1/agents/{agent_id}
PATCH  /v1/agents/{agent_id}

POST   /v1/tasks
GET    /v1/tasks?agent_id={agent_id}&status=in_progress&limit=25&cursor=OPAQUE_CURSOR
GET    /v1/tasks/{task_id}?include=artifacts
POST   /v1/tasks/{task_id}/state/operations
GET    /v1/tasks/{task_id}/events?after_sequence=0&limit=25
POST   /v1/tasks/{task_id}/checkpoints
GET    /v1/tasks/{task_id}/checkpoints
GET    /v1/tasks/{task_id}/checkpoints/{checkpoint_id}
POST   /v1/tasks/{task_id}/resume
POST   /v1/tasks/{task_id}/artifacts/upload
GET    /v1/tasks/{task_id}/artifacts/{artifact_id}/content
POST   /v1/tasks/{task_id}/artifacts
DELETE /v1/tasks/{task_id}/artifacts/{artifact_id}

Update state safely

POST /v1/tasks/tsk_.../state/operations
Authorization: Bearer $ELTEX_API_KEY
Idempotency-Key: pricing-progress-1
If-Match: "1"
Content-Type: application/json

{"operations":[
  {"op":"set_status","status":"in_progress"},
  {"op":"set_current_step","step":{"id":"collect","summary":"Collect pricing pages"}},
  {"op":"set_next_actions","items":["Compare plans"]}
]}

A request accepts 1–100 operations and resulting state is limited to 256 KiB. Successful updates return a new version, event_sequence, checksum, and ETag. A stale version returns 409 state_version_conflict; an omitted version returns 428 state_version_required.

Lists use opaque next_cursor values; events use next_after_sequence. Resume creates a new run from a verified checkpoint without erasing later history. Raw file uploads are stored privately in Cloudflare R2 and downloaded through an owner-authenticated endpoint; external artifact references can use https, gs, or s3 URIs.

Download complete Agent State reference

09 · AGENT RUNGUARD API

Authorize agent actions within explicit limits.

RunGuard stores an immutable policy snapshot per run, reserves estimated cost before an action, tracks call/runtime limits, and detects repeated actions, results, errors, and failures.

POST  /v1/guard-policies
GET   /v1/guard-policies
GET   /v1/guard-policies/{policy_id}
PATCH /v1/guard-policies/{policy_id}

POST  /v1/runs
GET   /v1/runs
GET   /v1/runs/{run_id}
GET   /v1/runs/{run_id}/events
POST  /v1/runs/{run_id}/events
POST  /v1/runs/{run_id}/preflight
POST  /v1/runs/{run_id}/executions/{execution_id}/complete
POST  /v1/runs/{run_id}/pause
POST  /v1/runs/{run_id}/resume
POST  /v1/runs/{run_id}/terminate
GET   /v1/runs/{run_id}/usage

Preflight before execution

POST /v1/runs/run_.../preflight
Authorization: Bearer $ELTEX_API_KEY
Idempotency-Key: run-1-model-call-7
Content-Type: application/json

{"action":{"action_type":"model_call","capability":"chat.completions","risk":"read_only","arguments":{"model":"eltex/smart"}},"estimated_cost_usd":0.04,"cost_source":"client_estimated"}

ALLOW and WARN include a short-lived execution token bound to the owner, run, execution, and exact action digest. PAUSE and TERMINATE do not. Complete the execution with the token, outcome, final cost, and optional Agent State/artifact progress evidence.

Download complete RunGuard reference

10 · AGENTIC WALLET & x402 API

Give one agent bounded USDC payment authority.

Agentic Wallet binds one agent identity, one dedicated ELTEX API key, and one immutable policy version. Policies control Base/Solana networks, per-transaction/daily/monthly USDC limits, autonomous approval thresholds, expiry, pause/revoke state, optional x402 domains, and optional direct-transfer recipient restrictions.

GET  /v1/agentic-wallet/balance?agent_id=agt_...
POST /v1/agentic-wallet/preflight
POST /v1/agentic-wallet/x402/fetch
POST /v1/agentic-wallet/transfer

Request an x402 resource

POST /v1/agentic-wallet/x402/fetch
Authorization: Bearer $ELTEX_API_KEY
Idempotency-Key: paid-resource-17
Content-Type: application/json

{
  "agentId": "agt_...",
  "network": "base",
  "url": "https://api.example.com/paid-resource",
  "method": "GET",
  "maxValue": "0.10"
}

An empty domain list allows all public HTTPS destinations. Adding exact or wildcard domains narrows access. ELTEX rejects credentials in URLs and local, private, link-local, multicast, or unresolved destinations. maxValue is a conservative USDC cap with at most six decimals.

Direct USDC transfer

POST /v1/agentic-wallet/transfer
Authorization: Bearer $ELTEX_API_KEY
Idempotency-Key: payout-17
Content-Type: application/json

{"agentId":"agt_...","network":"base","to":"0x1111111111111111111111111111111111111111","amount":"0.10"}

Direct transfers require permissions.transfer=true and a valid to/destination address on the selected network. An empty allowedRecipients list allows any valid address supplied at send time; a non-empty list restricts the destination to those exact entries. If approvalRequired is true, the owner approves the exact action under Agentic Wallet → Approvals. Retry the identical request with the identical idempotency key. Individual authorizations expire after ten minutes; indeterminate provider submissions remain submitting to prevent duplicate payment.

Download complete Agentic Wallet reference

11 · ERRORS & RETRIES

Use error.code, not message text.

{
  "error": {
    "message": "Your current capacity is already full",
    "type": "capacity_already_full",
    "code": "capacity_already_full"
  }
}

Agent Plan chat uses per-plan concurrency and has no separate documented per-minute quota. Elixir generation defaults to 10 requests/minute; read requests use the key limit, normally 60/minute. Use bounded exponential backoff for transient 5xx failures. Chat requests are not idempotent.

12 · OAUTH 2.0

Tokens, Device Flow, and OIDC status.

Register a public client

POST /oauth/clients
Authorization: Bearer <firebase-user-token>
Content-Type: application/json

{
  "name": "My App",
  "redirect_uris": ["https://app.example.com/oauth/callback"],
  "is_public": true
}

The endpoint returns 201 Created. The abbreviated response below matches the implemented envelope; public clients do not have a client_secret field.

{
  "ok": true,
  "client": {
    "id": "elx_client_...",
    "name": "My App",
    "is_public": true,
    "require_pkce": true
  }
}

Authorization Code + PKCE

GET /oauth/authorize
  ?response_type=code
  &client_id=elx_client_...
  &redirect_uri=https://app.example.com/oauth/callback
  &scope=openid profile email offline_access ai.chat:write
  &state=RANDOM_STATE
  &code_challenge=S256_CHALLENGE
  &code_challenge_method=S256

POST /oauth/token
{
  "grant_type": "authorization_code",
  "code": "elx_code_...",
  "redirect_uri": "https://app.example.com/oauth/callback",
  "client_id": "elx_client_...",
  "code_verifier": "ORIGINAL_VERIFIER"
}

{
  "access_token": "elx_at_...",
  "token_type": "Bearer",
  "expires_in": 3600,
  "refresh_token": "elx_rt_...",
  "scope": "openid profile email offline_access ai.chat:write"
}

refresh_token is present only when offline_access was requested and approved. Refresh with POST /oauth/token using grant_type: "refresh_token", refresh_token, and client_id. Refresh tokens rotate on every successful use.

Device Flow polling

POST /oauth/device/code
{"client_id":"elx_client_...","scope":"openid profile offline_access ai.chat:write"}

{
  "device_code": "elx_dev_...",
  "user_code": "ELTX-ABCD-EFGH",
  "verification_uri": "https://eltexlabs.com/device",
  "verification_uri_complete": "https://eltexlabs.com/device?user_code=ELTX-ABCD-EFGH",
  "expires_in": 900,
  "interval": 5
}

authorization_pending: continue polling at the interval. access_denied, expired_token, or invalid_grant: stop and restart only when appropriate.

Userinfo, revocation, and Connected Apps

GET  /oauth/userinfo
POST /oauth/revoke
GET  /account/connected-apps
DELETE /account/connected-apps/{client_id}
POST /account/connected-apps/{client_id}/revoke

/oauth/userinfo always requires a valid OAuth access token and returns sub. The profile and email scopes add their respective fields. Revocation accepts {"token":"elx_at_... or elx_rt_..."}.

OAuth error transport

{
  "error": {
    "message": "invalid_grant",
    "type": "invalid_request_error",
    "code": "invalid_request_error"
  }
}

For most OAuth endpoints, branch on the identifier in error.message. Current identifiers include unsupported_response_type, invalid_client, invalid_redirect_uri, pkce_required, invalid_pkce, invalid_grant, missing_params, unsupported_grant_type, and invalid_token. Device polling instead returns a top-level error such as {"error":"authorization_pending"}.

Agent Plan scopes are ai.chat:write, ai.models:read, ai.usage:read, and ai.capacity:write. Agent State uses agents:read/write and tasks:read/write. RunGuard uses runs:read, runs:write, and runtime.usage:read. Agentic Wallet uses dedicated API-key scopes wallet:read and wallet:spend; wallet routes reject OAuth tokens. Open the dashboard API Docs for complete mappings, refresh, revocation, and Elixir scopes.

13 · ELIXIR MEDIA API

Generation operations and production caveats.

Base URL: https://eltexlabs.com/v1/elixir

GET  /models
GET  /balance
GET  /generations?limit=50&cursor=OPAQUE_CURSOR
POST /images/generations
POST /images/edits
POST /video/generations
GET  /video/generations/{generation_id}
DELETE /video/generations/{generation_id}

Generation requests accept a stable, portable Idempotency-Key. Cursor pages contain data, has_more, and next_cursor; pass the opaque cursor back unchanged. Limits range from 1 to 100.

Generated asset URLs may be provider-hosted and may expire. No minimum retention window is guaranteed; copy completed assets promptly to storage you control.

Download Elixir API reference Download Elixir agent skill