Skip to main content
Glama
README.md
# Snap Router — MCP server

<!-- mcp-name: io.github.trigeochiral/snap-router -->

Task-to-tool discovery for AI agents, pay-per-call in USDC on Base (x402), free tier, no signup.

Hosted API: `https://trigeochiral.com` — set it as `SNAP_API`.

## Tool

| Tool | What |
|---|---|
| `snap_router` | Your task → the ranked x402 / MCP tools that serve it, over the merged x402 Bazaar + MCP Registry catalog (16k+). One pass, ~200ms, no LLM call. $0.003 |

## Why

Past ~2,000 tools, semantic retrieval beats keyword and beats long-context. Coinbase's Bazaar
`/ask` is LLM-ranked (a model call on every task); 402index falls back to keyword-only when its
embedding service times out. Snap Router has neither hole.

## Install

```bash
pip install snap-router-mcp
claude mcp add snap-router --env SNAP_API=https://trigeochiral.com -- snap-router-mcp
```

or with `uvx` (no install):

```bash
claude mcp add snap-router --env SNAP_API=https://trigeochiral.com -- uvx snap-router-mcp
```

Optional: `SNAP_XPAYMENT` = a base64 x402 `X-PAYMENT` payload, to make paid calls past the free
tier. Without it you get 25 free calls/day per IP.

## Direct HTTP (no MCP)

```bash
curl -sX POST https://trigeochiral.com/v1/snap -H 'content-type: application/json' \
  -d '{"task":"check a base token for honeypot before trading","k":4}'
```

## License

MIT

TDQS

A4.4/5.0

Scored across 1 tool

Disambiguation5/5

With only one tool, there is no possibility of misselection or overlap; its stated purpose (catalog discovery before loading tools) is unambiguous. An agent cannot confuse it with anything else in this server.

Naming Consistency5/5

The single tool follows a clean snake_case verb/resource-style name (snap_router) that matches the server name. There is no second convention to conflict with.

Tool Count3/5

A one-tool surface is inherently thin, even though a discovery router is arguably a single-purpose operation. The rubric treats 1-2 tools as borderline, and this sits right at that edge.

Completeness3/5

For a pure discovery role, the tool covers the core 'find candidates' operation, but the surface is a dead end: no way to fetch full details for a given service/tool ID, filter by price/category, or paginate the 16k-entry catalog. Agents can work around this by rerunning searches, but it is a notable gap.

Maintenance

ActivitySlowing
ResponsivenessNo issues