# RHF Impact MCP agent guide

This is the complete client-side guide for the RHF Impact MCP. It is safe to
save in a trusted agent-instructions location. It contains no API key or other
credential.

## Connect

| Setting | Value |
| --- | --- |
| MCP name | `impact` |
| URL | `https://mcp.researchhub.foundation/mcp` |
| Transport | Streamable HTTP |
| Authentication | `Authorization: Bearer <personal imcp key>` |
| Access | Read-only |

Keep the personal `imcp_…` key only in the agent host's secret environment,
using the variable name `IMPACT_MCP_API_KEY`. Never put a raw key in chat,
instructions, source control, or this file.

### Codex configuration

```toml
[mcp_servers.impact]
url = "https://mcp.researchhub.foundation/mcp"
bearer_token_env_var = "IMPACT_MCP_API_KEY"
enabled_tools = [
  "get_impact_catalog",
  "search_impact_records",
  "get_impact_record",
]
```

The three entries above are the complete MCP API. The workflow library below
is retrieved through those tools; workflows are not separate shell tools or
browser extensions.

## Use the MCP

1. At the beginning of each research task, call `get_impact_catalog`.
2. Select the relevant workflow from `workflow_skills.records` and retrieve
   its complete guidance with `get_impact_record` and the listed `document_id`.
3. Use `search_impact_records` only to find candidate records.
4. Retrieve every record that supports a material finding with
   `get_impact_record`.
5. Preserve the returned `document_id`, source, corpus version, evidence, and
   uncertainty in the final work.

### MCP tools

| Tool | Use |
| --- | --- |
| `get_impact_catalog` | Discover the active corpus and its complete live workflow list. |
| `search_impact_records` | Find candidate funder, project, researcher, profile, NIH-project, or skill records. |
| `get_impact_record` | Retrieve one complete published record by its stable `document_id`. |

## Workflow library

The active catalogue is authoritative: it determines which workflows are
available at the time of use. The current RHF Impact workflow library is:

| Workflow | Use |
| --- | --- |
| `find-project-expert` | Identify evidence-backed subject-matter experts for a project or research need. |
| `find-project-funder` | Identify evidence-backed potential funders for a project or research need. |
| `funder-cofunding-map` | Map funders that co-appear on published project evidence. |
| `funder-discovery` | Build a public-evidence funder prospect shortlist. |
| `funder-due-diligence` | Create a bounded public-record profile of a potential funder. |
| `funder-fit-scorecard` | Score a shortlist against explicit, evidence-backed fit criteria. |
| `funder-peer-benchmark` | Compare named funders on defined coverage criteria. |
| `funder-portfolio-landscape` | Map a recorded funder portfolio by theme, mechanism, recipient, and geography. |
| `funder-white-space` | Surface apparent evidence gaps for a defined strategy question. |
| `funding-case-report` | Produce an evidence-conscious decision case report. |
| `grant-opportunity-brief` | Turn a funding need into a prospect brief before external validation. |
| `impact-crossref-linkage` | Link grant activity to publication metadata from Crossref. |
| `impact-nih-mining` | Mine NIH RePORTER grants for funding-impact linkage. |
| `impact-nsf-mining` | Mine NSF awards and outcomes metadata for impact linkage. |
| `impact-quality-gates` | Apply quality checks before snapshot publishing. |
| `impact-results-export` | Create clean end-user results exports and reports. |
| `impact-survey-report` | Create descriptive stakeholder reports from supplied survey responses. |
| `project-precedent-scan` | Find comparable projects, awards, and approaches. |
| `proposal-readiness` | Assess evidence-backed proposal readiness and preparation gaps. |
| `research-team-design` | Design a proposal team matrix from published expertise and project context. |
| `researcher-discovery` | Find and compare published researcher profiles. |
| `researchhub-report-generation` | Create source-backed ResearchHub research and funding reports. |

## Rules

- Treat search results as discovery leads, not conclusions or rankings.
- Treat historical funding, field overlap, and co-appearance as diligence
  signals—not current funding, eligibility, availability, endorsement, or a
  live opportunity.
- Verify current deadlines, programme facts, legal or financial facts, and
  funding availability against primary sources.
- Do not access, reveal, retain, or request credentials, infrastructure,
  databases, local files, or unapproved external sources.
- Do not make paid, publishing, promotional, or other mutating actions without
  explicit user approval.

## Starter instruction

```text
Use only the configured RHF Impact MCP tools. Start with get_impact_catalog.
For every substantive task, select the relevant workflow from
workflow_skills.records, retrieve it with get_impact_record, then search for
candidate records and retrieve every record supporting a material finding.
Keep document IDs, source provenance, corpus version, uncertainty, and needed
primary-source follow-up in the response. Never access credentials,
infrastructure, local files, or unapproved external sources, and never perform
paid or mutating actions without explicit approval.
```