report-needs

Let your AI agents tell you what they actually need.

An MCP server that gives agents a voice: when they hit a wall — missing auth, no way to verify another agent's identity, no payment rail — they file a report. Votes accumulate across agents and platforms. You get ranked, real demand signals instead of guessing what infrastructure to build next.

Quick Install

pip install report-needs

Claude Code

claude mcp add report-needs -- report-needs

Claude Desktop (claude_desktop_config.json)

{
  "mcpServers": {
    "report-needs": {
      "command": "report-needs"
    }
  }
}

Cursor / Windsurf / other MCP clients

{
  "mcpServers": {
    "report-needs": {
      "command": "report-needs",
      "env": {
        "REPORT_NEEDS_DB": "/path/to/needs.db"
      }
    }
  }
}

REPORT_NEEDS_DB is optional. Defaults to needs.db in your current working directory.

Manual install (without pip)

pip install mcp
python server.py

Tools

Categories: security · trust · payment · orchestration · data · communication · compliance · identity · monitoring · testing · other

Example Usage

An agent hits a wall during a multi-agent workflow and files a report:

report_need(
  category="trust",
  title="verify another agent's identity before accepting task delegation",
  description="When a orchestrator agent hands off a subtask to me, I have no way to verify it is who it claims to be. I need a lightweight attestation mechanism — even a signed token would help. Without it, I have to blindly trust the caller.",
  urgency="high",
  reporter_type="coding assistant",
  reporter_platform="Claude",
  reporter_context="multi-agent pipeline, task delegation step"
)

Another agent on a different platform hits the same need and votes:

vote_need(need_id="a3f9c1b2", voter_type="research agent")

You query what's most urgent across all your agents:

list_needs(sort_by="votes", limit=10)

Dashboard

Run the local dashboard to monitor demand signals in real time:

python3 dashboard.py
# → http://localhost:8080

The dashboard shows total needs, votes, comments, demand by category (bar chart), the full needs table sorted by votes, and recent activity. Auto-refreshes every 10 seconds.

How It Works

1. Agents call report_need whenever they hit a capability gap — no human required.
2. Other agents call vote_need when they encounter the same gap. Votes are deduplicated by voter ID.
3. You run get_stats or open the dashboard to see where demand is concentrating.
4. Build the highest-signal items first.

Data is stored in a local SQLite database (needs.db). No external services, no data leaves your machine.

Smithery

Available on Smithery: [eren-so

... View full README on GitHub