# Titan Frameworks MCP

> Live, token-efficient documentation tools for fast-evolving AI frameworks.

Titan Frameworks is a Model Context Protocol server. Agents call tools over stdio (Cursor/Claude Desktop) or Streamable HTTP (`POST /mcp`).

**Production MCP:** `https://titan-frameworks-production.up.railway.app/mcp`  
Production HTTP `/mcp` is billed with x402 at **$0.001 USDC on Base** or **1000 drops XRP / 0.001 RLUSD** on XRPL mainnet.

## Endpoints

| Method | Path | Price | Auth |
| --- | --- | --- | --- |
| GET | `/health` | free | none |
| GET | `/llms.txt` | free | none |
| GET | `/tools` | free | none |
| GET/DELETE | `/mcp` | free | none (Streamable HTTP session) |
| POST | `/mcp` `initialize` / `ping` / `tools/list` | free | none |
| POST | `/mcp` `tools/call` | $0.001 USDC on Base (`eip155:8453`) **or** 1000 drops XRP / RLUSD on XRPL mainnet (`xrpl:0`) | x402 `PAYMENT-SIGNATURE` / `X-PAYMENT` / `X-Payment-Signature` |
| GET | `/admin` | private (password login, not a public dashboard) | `ADMIN_SECRET_KEY`. Live: `https://titan-frameworks-production.up.railway.app/admin`. Old `/admin/api/stats` is 404. |

Local stdio mode (`npx tsx src/index.ts`) does not charge.

## Pricing

- **Tool calls over HTTP `/mcp`:** `0.001` USDC on Base **or** `1000` drops XRP (`0.001` XRP) / RLUSD on XRPL mainnet
- **Networks:** Base mainnet (`eip155:8453`) and XRPL mainnet (`xrpl:0`)
- **Assets:** USDC `0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913`, native `XRP`, RLUSD (`524C555344…` + issuer)
- **Scheme:** `exact` (x402)
- **Production:** billing is live (`X402_ENABLED=true`, `X402_USDC_LIVE=true`, `X402_XRPL_LIVE=true`). Missing payment headers on `tools/call` return HTTP 402 with machine-readable `accepts[]` plus PayAI bazaar / x402scan discovery metadata. Handshake methods stay unpaid.
- **Live Base USDC:** PayAI facilitator `https://facilitator.payai.network`.
- **Live XRPL:** T54 facilitator `https://xrpl-facilitator-mainnet.t54.ai` (`XRPL_NETWORK=xrpl:0`).

## Tools (JSON Schema)

### search_ai_framework_docs

Keyword search over a local Markdown index. Use for concepts or symbols; `hitCount` 0 means no match. Not for stack traces (`diagnose_framework_error`) or copy-paste syntax (`fetch_latest_syntax`). Paid `tools/call`.

```json
{
  "name": "search_ai_framework_docs",
  "input": {
    "type": "object",
    "required": ["framework", "query"],
    "properties": {
      "framework": { "type": "string", "enum": ["langchain", "llamaindex", "ollama", "xrpl"] },
      "query": { "type": "string" }
    }
  },
  "output": {
    "type": "object",
    "required": ["framework", "query", "hitCount", "hits"],
    "properties": {
      "hitCount": { "type": "integer" },
      "hits": {
        "type": "array",
        "items": {
          "type": "object",
          "properties": {
            "title": { "type": "string" },
            "url": { "type": "string" },
            "score": { "type": "number" },
            "chunk": { "type": "string" }
          }
        }
      }
    }
  }
}
```

### fetch_latest_syntax

Curated snippets and migration notes for one topic. Use when writing or migrating code. Unknown topics fall back to related doc chunks. Paid `tools/call`.

```json
{
  "name": "fetch_latest_syntax",
  "input": {
    "type": "object",
    "required": ["framework", "topic"],
    "properties": {
      "framework": { "type": "string", "enum": ["langchain", "llamaindex", "ollama", "xrpl"] },
      "topic": { "type": "string", "description": "agents | rag | chat | tools | streaming | migration | embeddings" }
    }
  }
}
```

### diagnose_framework_error

Match a stack trace against known deprecated APIs. Use only when you have an `error_log`; `matched=false` means search next. Paid `tools/call`.

```json
{
  "name": "diagnose_framework_error",
  "input": {
    "type": "object",
    "required": ["framework", "error_log"],
    "properties": {
      "framework": { "type": "string", "enum": ["langchain", "llamaindex", "ollama", "xrpl"] },
      "error_log": { "type": "string" }
    }
  }
}
```

## Indexed frameworks

- LangChain (`langchain`) — Python/JS LCEL, agents, LangGraph notes
- LlamaIndex (`llamaindex`) — indexes, Settings, query/chat engines
- Ollama (`ollama`) — CLI, REST `/api/chat`, Python/JS SDK
- XRPL (`xrpl`) — xrpl.js / xrpl-py Client, Wallet, Payments, channels, trustlines, RLUSD, Hooks notes

Refresh: nightly in production when the index is older than 24 hours. Local: `npm run index-docs`.

## Agent notes

Keep tool arguments short. Responses are already truncated for context windows (top hits, ~1.2k chars/chunk). Prefer `fetch_latest_syntax` before pasting large docs into a prompt.