Secant Agent Research
# Secant Agent Research MCP
[](https://github.com/jnilrac/secant-agent-research-mcp/actions/workflows/ci.yml)
[](https://glama.ai/mcp/servers/jnilrac/secant-agent-research-mcp)
[](https://glama.ai/mcp/servers/jnilrac/secant-agent-research-mcp)
[](https://smithery.ai/servers/hello-j2mo/secant-agent-research-pack)
MCP wrapper for the [Secant Agent Research Pack](https://agentic.secantoutreach.com/agent-research), a paid x402 research API for autonomous agents.
The server exposes tools for:
- `search`: ranked web search.
- `research_pack`: search, selected page extraction, normalized JSON, and citations.
- `extract_page`: readable page text and metadata extraction.
- `monitor_diff`: page change checks.
## Tool Behavior
All tools use the same retry-safe x402 pattern:
1. Call the tool without `payment` to receive the upstream payment requirement.
2. Pay the quoted Base USDC requirement with an x402-compatible client.
3. Retry with the payment proof and, when available, the same `payment_identifier`.
The wrapper returns `payment_required: true` and the upstream 402 body for unpaid calls. It does not execute paid work until the live API accepts the payment proof.
The native paid API uses HTTP `402` and Base USDC over x402. If you call a tool without a payment proof, the tool returns the upstream payment requirement so an agent can pay and retry. If you include a valid payment proof, the tool forwards it to the live API.
## Run
```bash
docker build -t secant-agent-research-mcp .
docker run --rm -i secant-agent-research-mcp
```
## Environment
| Variable | Default | Description |
| --- | --- | --- |
| `SECANT_AGENT_RESEARCH_API_BASE_URL` | `https://agentic.secantoutreach.com` | Base URL for the live Secant Agent Research API. |
## Canonical Discovery
- Buyer page: <https://agentic.secantoutreach.com/agent-research>
- Smithery listing: <https://smithery.ai/servers/hello-j2mo/secant-agent-research-pack>
- x402 manifest: <https://agentic.secantoutreach.com/.well-known/x402.json>
- OpenAPI: <https://agentic.secantoutreach.com/openapi.yaml>
- Agent card: <https://agentic.secantoutreach.com/agent-card.json>
TDQS
Scored across 4 tools
The tools are distinct in purpose: search for web searching, extract_page for content from known URLs, monitor_diff for change detection, and research_pack as a higher-level combined workflow. While research_pack encompasses search and extraction, it is presented as a different service level, so confusion is minimal.
Three tools follow a verb_noun pattern (extract_page, monitor_diff, research_pack), with 'research_pack' being noun_noun but still clear. 'search' is a single verb, slightly deviating but acceptable. Overall pattern is recognizable and consistent.
Four tools is an appropriate number for a research-focused server, covering key operations (search, extraction, monitoring, and a bundled paid workflow) without being overbearing or insufficient.
Core research workflows (search, extract, monitor, full research pack) are covered. Minor gaps exist, such as lacking a tool to list or retrieve previous research packs, but the set is functional for autonomous agents.