# I Ninte Grad Agent MCP server

Source: https://glimind.com/server/mcp-registry%2Fio.github.i-ninte%2Fgrad-agent
Registry id: `mcp-registry/io.github.i-ninte/grad-agent`
Status: unknown

## Install

```json
{
  "mcpServers": {
    "i-ninte-grad-agent-mcp-server": {
      "command": "uvx",
      "args": [
        "grad-agent"
      ]
    }
  }
}
```

Claude Code:

```sh
claude mcp add i-ninte-grad-agent-mcp-server -- uvx grad-agent
```

## Tools (22)

- `discover_profs`: arXiv + OpenReview scan (raw candidates, no verification).
- `draft_cold_email`: Manual per-prof draft.
- `draft_sop`: Compile a Columbia-style SOP PDF (versioned).
- `followups_due`: Drafted nudges for profs silent 10+ days.
- `ingest_replies`: Read-only IMAP scan; auto-tags replies in the log.
- `interview_prep`: One-page brief: papers, likely questions, talking points.
- `list_programs`: Your target programs.
- `list_projects_in_catalog`: Show every project the matcher can see.
- `list_scholarships`: External scholarships filtered by eligibility region.
- `outcome_report`: Response rates by area and project.
- `outreach_log_view`: Show last N rows of the outreach xlsx.
- `outreach_mark_followup`: Record a sent nudge (never nudged twice).
- `outreach_mark_response`: Tag replies; feeds the matcher's learning loop.
- `outreach_mark_sent`: Flag a row as actually sent to the prof.
- `run_daily_batch`: Full pipeline: verify, recruit, hook, verify, draft, log.
- `run_program_batch`: Deadline-driven batch: draft for one program's faculty.
- `s2_cache_invalidate`: Selectively purge Semantic Scholar cache entries.
- `send_draft_to_me`: Ship any draft file to your review inbox.
- `skipped_log_view`: Audit trail of skipped leads with stage and mismatch reason.
- `sync_catalog`: Pull projects from github, hf, or local.
- `upcoming_deadlines`: Any program deadline in the next N days.
- `upcoming_scholarship_deadlines`: Scholarship deadlines approaching.

Measured by Glimind (https://glimind.com): MCP handshake + tool list only, no tool is ever called. CC BY 4.0.
