Arcade Engine API
Tools that enable LLMs to interact directly with the engine API.
0.4.0Arcade Engine API Toolkit
Provides LLMs with direct programmatic access to the Arcade Engine API, enabling agents to manage tools, workers, authentication, secrets, projects, and executions at runtime.
Capabilities
- Tool discovery & execution: List, inspect, and execute any tool available in the engine; retrieve OpenAPI specs, formatted specs, and paginated tool lists filtered by toolkit or provider.
- Execution history & auditing: Page through tool run history with rich filtering (tool name, toolkit, user, outcome, time range); retrieve full execution records including optional inputs/outputs (requires project admin authority, is audited).
- Worker lifecycle management: Create, retrieve, update, test, and delete workers; check worker health status and authorize workers by ID.
- Authentication & connection management: List, inspect, create, and delete auth providers; manage user/auth-provider connections; authorize users for specific tools and monitor ongoing authorization status.
- Secrets management: Create and delete secrets by ID within the engine.
- Project & engine operations: List accessible projects, retrieve project details, update session verification settings, and check overall engine health.
Secrets
ARCADE_API_KEY — An API key that authenticates requests from this toolkit to the Arcade Engine API. To obtain one, sign in to the Arcade dashboard, navigate to Auth → Secrets, and generate a new API key. The key must have sufficient permissions to perform the operations your agent requires (e.g., project admin authority is required to retrieve tool execution inputs/outputs).
For guidance on configuring secrets in Arcade tools, see the Arcade secrets documentation. You can also manage secrets directly at https://api.arcade.dev/dashboard/auth/secrets.
Available tools(29)
| Tool name | Description | Secrets | |
|---|---|---|---|
Authorize a user to access a specific tool.
This tool authorizes a user for a specific tool by its name. It should be called when a user needs permission to access a specific tool. The tool returns the authorization status. | 1 | ||
Authorize a worker based on their ID.
This tool is used to verify and authorize a worker by their ID. It should be called when it's necessary to check if a worker has the appropriate permissions or status to perform certain tasks. | 1 | ||
Check the health status of the Arcade Engine.
Use this tool to verify if the Arcade Engine service is currently healthy and operational. | 1 | ||
Verify the ongoing authorization status of a tool.
Use this tool to check the status of an ongoing authorization process for a specific tool. Ideal for monitoring when an authorization completes or times out. | 1 | ||
Create a new worker in the system.
This tool is used to add a new worker to the system. It should be called when there's a need to register a new worker. | 1 | ||
Delete a specific auth provider by ID.
This tool deletes a specified authentication provider using its ID. It should be called when you need to remove an auth provider from the system. | 1 | ||
Delete the Model Context Protocol endpoint data.
This tool deletes data at the Model Context Protocol (MCP) endpoint, which supports Streamable HTTP transport. Use it to remove existing configurations or data tied to this endpoint. | 1 | ||
Deletes a secret using its unique ID.
Use this tool to delete a specific secret identified by its ID. Useful when needing to permanently remove a secret from the system. | 1 | ||
Deletes a user/auth provider connection.
This tool deletes a connection between a user and an authentication provider. It should be called when a user needs to disconnect their account from an external auth provider. | 1 | ||
Deletes a specified worker from the system.
Use this tool to remove a worker by providing their unique ID. It should be called when you need to permanently delete a worker record from the system. | 1 | ||
Execute a specified tool with given parameters.
This tool allows the execution of a specified tool by providing its name and necessary arguments. It's useful for triggering specific actions or processes as dictated by the tool's logic.
Note: Understanding the request schema is necessary to properly create
the stringified JSON input object for execution.
Modes:
- GET_REQUEST_SCHEMA: Returns the schema. Only call if you don't
already have it. Do NOT call repeatedly if you already received
the schema.
- EXECUTE: Performs the operation with the provided request body
JSON.
If you need the schema, call with mode='get_request_schema' ONCE, then execute. | 1 | ||
Retrieve a list of tools for a specific worker.
This tool is used to get a list of tools associated with a specific worker ID. It should be called when you need an overview of tools for a particular worker. | 1 | ||
Retrieve details of a specific authentication provider.
Use this tool to obtain the details of a particular authentication provider by specifying its ID. | 1 | ||
Fetches a formatted specification for a given tool.
Use this to obtain detailed, formatted specifications for a specific tool from a provider. | 1 | ||
Get the OpenAPI 3.0 specification in JSON format.
Use this tool to retrieve the OpenAPI 3.0 specification for the Arcade Engine API, which provides detailed information about all available endpoints and their usage. | 1 | ||
Retrieve detailed information about a specific project.
This tool fetches and returns detailed information for a given project based on its ID. Use it when you need to access specific details about a project. | 1 | ||
Retrieve the details for a single tool execution.
Use this tool to get the full record of one tool run by its execution ID, including each attempt the engine made. The recorded tool inputs and outputs are withheld by default because they can be large and carry sensitive data; set include_inputs or include_outputs to true to request them, which requires project admin authority and is audited. Returns a 404 error when the project holds no execution with that ID. This replaces the retired scheduled_tools detail endpoint and covers both scheduled and immediate runs. | 1 | ||
Retrieve a list of tools from the engine configuration.
Use this tool to get a paginated list of tools available in the engine configuration, with optional filtering by toolkit. | 1 | ||
Retrieve the specification for a specific arcade tool.
This tool returns the arcade tool specification for a specified tool by its name. It should be called when there's a need to understand the details or capabilities of a particular tool. | 1 | ||
Retrieve worker details using their ID.
Use this tool to obtain detailed information about a worker by providing their unique ID. Ideal for situations where specific worker data is required. | 1 | ||
Retrieve the health status of a worker.
This tool is used to check the health status of a specific worker by their ID. It should be called when you need to monitor or verify the operational status of a worker. | 1 | ||
Retrieve a list of accessible projects.
This tool returns all projects that the caller has access to. It is useful for identifying projects available to a specific user or account. | 1 | ||
Retrieve all authentication connections for users.
Use this tool to get a comprehensive list of all authentication connections associated with users. Ideal for managing or auditing user authentication setups. | 1 | ||
Retrieve a list of available authentication providers.
This tool fetches a paginated list of authentication providers accessible to the caller. It should be used when identifying or managing auth providers in the system. | 1 | ||
List the tool executions recorded for a project.
Use this tool to page through the history of tool runs in a project, optionally filtered by tool name, toolkit, user, outcome, failure text, or time range. This replaces the retired scheduled_tools list endpoint and covers both scheduled and immediate runs. There is no server-side filter for the kind of run: every returned row carries an execution_type field whose value is 'immediate' or 'scheduled', so select on that field yourself when you only want scheduled executions. Rows describe what ran, when, and how it went; they never carry the recorded tool inputs or outputs, which are available from GetToolExecutionDetails. | 1 |