MCPolly

What is MCPolly?

MCPolly is an MCP (Model Context Protocol) server that allows AI agents to report their progress, errors, and state in real time. It provides a unified web dashboard for humans to monitor all agents, and includes a knowledge layer for indexing and semantically searching product documents. Built with Rust, Axum, SQLite, and HTMX, it is designed for self-hosting on minimal hardware.

How to use MCPolly?

Install via the install script or build from source, then start the server with PORT=3000 RUST_LOG=mcpolly=info ./mcpolly. On first run, a setup wizard in the browser provides a default API key. Configure your AI agent platform (e.g., Cursor, Claude Code) to connect to the server’s /mcp endpoint using HTTP Bearer auth or the optional stdio bridge binary. Agents can then use MCP tools like register_agent, post_status, and post_error to report activity.

Key features of MCPolly

• Agent registration and real‑time status tracking (starting, running, error, completed, etc.)
• Real‑time activity feed and global error feed across all agents
• Configurable webhook alerts (Discord, Slack, generic) with retry logic
• Dark‑mode web dashboard with live updates via HTMX polling
• Vector semantic search over indexed documents using local Ollama embeddings
• Agent spawning with contextual knowledge from indexed content

Use cases of MCPolly

• Monitor multiple AI agents working on different tasks from a single dashboard
• Automatically index product requirements, design docs, and reference materials for semantic search
• Get instant alerts when an agent encounters an error or completes a critical task
• Enable agents to self‑report progress and state without human prompting
• Self‑host a lightweight observability platform for agent‑driven workflows

FAQ from MCPolly

How does MCPolly communicate with AI agents?

MCPolly exposes an MCP Streamable HTTP endpoint (/mcp) using JSON‑RPC over HTTP with SSE streaming. Agents connect directly via HTTP; an optional stdio binary (mcpolly_mcp) is available for platforms that do not support HTTP transport.

Does MCPolly require Ollama to work?

No, Ollama is optional. Vector embeddings for semantic search and agent spawning with context require Ollama running locally with the all‑minilm model. Without it, the web dashboard and agent observability still function.

How do I get or reset my API key?

On first run, the server prints a default API key and opens a setup wizard. If you forget the key, click “Forgot API key? Reset your instance” on the login page to revoke all keys, generate a new one, and re‑enter the setup wizard.

Where is data stored?

All agent data, activity logs, error records, and vector embeddings are stored in a local SQLite database (mcpolly.db) created automatically on first run. No external database is required.

What authentication does MCPolly use?

MCPolly uses Bearer token authentication. The API key is passed in the Authorization header of HTTP requests. The server supports multiple API keys managed via the web dashboard’s settings page.