romankurnovskii/apina/apina-api-registry

Apina API Registry Skill

Use this skill to interface with the Apina API registry. Apina serves as a centralized lookup engine containing OpenAPI schemas of registered Web3 and REST providers (such as DEX Screener, Meteora DLMM, etc.).

1. Agent Workflow: How to Use Apina (Core Instructions)

When you are asked to query or integrate an external API:

1. Never Guess: Do not make assumptions about the endpoint paths, parameter names, or base URLs of target services.
2. Search the Registry: Use the apina MCP tool search_api or query the REST endpoints to locate the target provider and endpoint.
3. Construct the Request: Use the retrieved schema (paths, query parameters, types) and the provider's baseUrl to construct the HTTP request.

Example Walkthrough for Agents

If the user asks: _"Check the liquidity pools on Meteora matching the query 'SOL'"_:

Step A: Search for the endpoint

Call the MCP tool search_api (or query GET /api/v1/providers/meteora/endpoints?search=pool):

• Tool Call: apina/search_api(query="pools", provider_id="meteora")
• Response Snippet:

``` Provider: Meteora DLMM API (meteora) Endpoint: GET /pools Summary: Pools Description: Returns a paginated list of pools Parameters:

• query (string, optional): Search query matching pools by tokens/address
• page_size (integer, optional): Max 1000

```

Step B: Fetch the Provider Base URL

Call the REST endpoint GET /api/v1/providers/meteora (or view config/providers.json):

• Response:

``json { "id": "meteora", "base_url": "https://dlmm.datapi.meteora.ag" } ``

Step C: Execute the Target Request

Now that you have the verified base URL and path parameters, construct and execute the call:

• Request: GET https://dlmm.datapi.meteora.ag/pools?query=SOL&page_size=10

2. API Reference for Agents

A. MCP Server Interface

If the apina MCP server is registered in your client, use the following tool:

search_api

• Arguments:
• query (string, required): A keyword to search endpoint paths, summaries, or descriptions (e.g. "pools", "pairs", "token").
• provider_id (string, optional): Filter search to a specific provider (e.g. "meteora", "dexscreener").
• Returns: Text block summarizing matching endpoints, including path, method, and query/path parameter requirements.

B. REST API Interface

If the REST API is running (typically at http://localhost:8081 or http://localhost:8000), you can query the endpoints directly via HTTP:

• List Providers: GET /api/v1/providers

Returns all registered providers with metadata, base URLs, and total endpoint counts.

• Get Provider Details: GET /api/v1/providers/{provider_id}

Returns the configuration (including base_url) for a specific provider.

• List Endpoints: GET /api/v1/providers/{provider_id}/endpoints

Filter and list endpoints using query parameters: search, method, tag, page, and limit.

• Get Endpoint Details: GET /api/v1/providers/{provider_id}/endpoints/{endpoint_id}

Returns the complete parameter schemas, responses, and description for a single endpoint.

3. Setup & Administration (For Humans)

To run the Apina service locally, you can choose between Docker Hub (recommended) or running python locally.

Option A: Running via Docker Hub

The Docker image `romankurnovskii/apina:latest` bundles the registry and schemas internally.

• Run the REST API Server:

``bash docker run -d --name apina-api -p 8081:8000 romankurnovskii/apina:latest ``

Verify via http://localhost:8081/health.

• Configure the MCP Server (stdio transport):

Add the following block to your client configuration (e.g. mcp_config.json):

``json { "mcpServers": { "apina": { "command": "docker", "args": ["run", "-i", "--rm", "romankurnovskii/apina:latest", "python", "service/mcp_server.py"] } } } ``

Option B: Running Locally with Python

1. Install dependencies:

``bash pip install -r service/requirements.txt ``

1. Start the REST API:

``bash cd service uvicorn main:app --host 0.0.0.0 --port 8000 ``

1. Register the local MCP Server:

``json { "mcpServers": { "apina": { "command": "python3", "args": ["/absolute/path/to/apina/service/mcp_server.py"] } } } ``