How do I generate agent docs with the Elba API?

Send a POST request to /api/generate with your API description, OpenAPI spec, or plain English. Elba will return a full AgentSpec with actions, capabilities, and documentation.

How do agents discover APIs on Elba?

Agents can discover APIs through multiple endpoints: GET /api/agent-search?q=query for search, GET /api/registry for the full listing, GET /api/agent/{id} for individual specs, and /.well-known/agent.json for the discovery manifest.

Does the Elba API support CORS?

Yes, all agent-facing endpoints (agent, registry, search, MCP, embed) include Access-Control-Allow-Origin: * headers for cross-origin access.

How do I get the MCP config for a spec?

Send a GET request to /api/mcp/{spec_id}. It returns an MCP server config with tools mapped from the spec actions, compatible with Claude Desktop and other MCP clients.

← Home

API Reference

Complete reference for all Elba endpoints. All agent-facing endpoints return JSON with CORS headers and are publicly accessible.

Base URL: https://useelba.com

Spec Management

Create, read, update, and manage your AgentSpecs.

POST/api/generate

Generate a new AgentSpec from raw text or OpenAPI input. Automatically extracts capabilities and reasoning docs.

Parameters

input (string) — API description, OpenAPI spec, or plain English

Response

AgentSpec object with id, actions, capabilities, and docs

GET/api/docsCORS

List all specs. Returns an array of AgentSpec objects sorted by creation date.

Response

AgentSpec[]

GET/api/spec/{id}CORS

Get a single spec by ID. Returns the full AgentSpec with all actions and metadata.

Parameters

id (path) — Spec ID

Response

AgentSpec object

POST/api/spec/{id}/update

Update spec name and description without regenerating.

Parameters

name (string, optional) — New name
description (string, optional) — New description

Response

Updated AgentSpec object

POST/api/spec/{id}/add-action

Generate and add a new action to an existing spec using AI.

Parameters

description (string) — What the new action should do

Response

{ action: AgentAction, spec: AgentSpec }

POST/api/spec/{id}/regenerate-action

Regenerate an existing action with improved documentation using AI.

Parameters

actionName (string) — Name of the action to regenerate

Response

{ action: AgentAction, spec: AgentSpec }

POST/api/spec/{id}/remove-action

Remove an action from a spec. Cannot remove the last action.

Parameters

actionName (string) — Name of the action to remove

Response

{ removed: string, spec: AgentSpec }

Publishing

Publish specs to custom subdomains.

POST/api/publish

Publish a spec to a custom subdomain. Validates subdomain uniqueness.

Parameters

id (string) — Spec ID
subdomain (string) — Desired subdomain
siteTitle (string, optional) — Site display title
brandColor (string, optional) — Hex color
isPublic (boolean, optional) — Visibility

Response

Updated AgentSpec object (409 if subdomain taken)

GET/api/subdomain-check

Check if a subdomain is available.

Parameters

subdomain (query) — Subdomain to check
excludeId (query, optional) — Spec ID to exclude

Response

{ available: boolean, subdomain: string }

Agent Discovery

Endpoints agents use to find and understand your API.

GET/.well-known/agent.jsonCORS

Well-known agent discovery endpoint. Lists all published specs with capabilities and discovery URLs.

Response

Agent discovery manifest with tools array

GET/api/agent/{id}CORS

Per-spec agent.json with full details. Tracks agent_fetch analytics events.

Parameters

id (path) — Spec ID

Response

Agent JSON payload with version, capabilities, actions, and discovery URLs

GET/api/mcp/{id}CORS

MCP server config for a spec. Each action becomes a typed MCP tool.

Parameters

id (path) — Spec ID

Response

MCP server config with tools and input schemas

GET/api/registryCORS

Full tool registry. Lists all specs with capabilities.

Response

{ tools: [{ name, description, capabilities, agent_json_url }] }

GET/api/agent-searchCORS

Search specs by capability. Weighted scoring by name (10x), capability (8x), description (5x), action (4x).

Parameters

q (query) — Search query

Response

{ tools: [{ name, description, capabilities, score, agent_json_url }] }

GET/llms.txt

LLM context file listing all APIs, endpoints, and capabilities. Follows the emerging llms.txt standard.

Response

Plain text

Embed & Analytics

Embed your docs anywhere and track usage.

GET/api/embed/{id}

JavaScript snippet that renders an embeddable iframe of your spec. Include as a script tag on any page.

Parameters

id (path) — Spec ID

Response

application/javascript

GET/api/analytics/{id}

Get analytics for a spec — views, agent fetches, search hits, MCP downloads.

Parameters

id (path) — Spec ID

Response

{ views, agent_fetches, search_hits, mcp_fetches, total, last_accessed }

GET/api/dashboard

Aggregated analytics across all specs.

Response

{ specs: [{ id, name, published, subdomain, analytics }] }

Playground

Execute prompts against a spec.

POST/api/run

Run a natural language prompt against a spec. Maps the prompt to an action and returns mock output.

Parameters

prompt (string) — Natural language prompt
specId (string) — Spec ID

Response

{ action: string, inputs: object, output: object }

POST/api/search

Human-facing search. Returns specs matching a query.

Parameters

query (string) — Search query

Response

{ results: [{ doc_id, name, description, capabilities, score }] }

API Reference

Spec Management

Publishing

Agent Discovery

Embed & Analytics

Playground

Related documentation