antientropy-mcp
# antientropy-mcp
MCP server for searching and reading articles from the [AntiEntropy Resource Portal](https://resourceportal.antientropy.org/docs). Gives Claude Code (or any MCP client) access to 140+ articles on nonprofit governance, compliance, HR policies, and fiscal sponsorship.
## Tools
- **antientropy_glob** — find articles by title/category path (`*policy*`, `Governance*/**`)
- **antientropy_grep** — regex search across article content, ripgrep-style output
- **antientropy_read** — read an article by slug with line numbers
- **antientropy_categories** — browse the full category tree
## Usage
### Remote (hosted on Fly.io)
```bash
claude mcp add --transport http antientropy https://antientropy-mcp.fly.dev/mcp
```
### Local (stdio)
```bash
claude mcp add antientropy -- uv run --directory /path/to/antientropy-mcp python -m antientropy_mcp
```
Populate the cache on first use:
```bash
uv run antientropy-sync
```
Articles are cached locally (~`~/.antientropy-mcp`). The remote server syncs daily via cron.
## Development
```bash
uv sync
uv run pytest
```
## Deploy
```bash
flyctl deploy
```
Requires a Fly.io volume named `data` mounted at `/data`. See `fly.toml` for config.
TDQS
Scored across 4 tools
Each tool has a clearly distinct purpose: antientropy_categories lists the category tree, antientropy_glob finds articles by path pattern, antientropy_grep searches article content with regex, and antientropy_read retrieves article content by slug. There is no overlap or ambiguity between these functions.
All tool names follow a consistent 'antientropy_' prefix with descriptive suffixes (categories, glob, grep, read) in snake_case. This pattern is uniform throughout the set, making the tools easily identifiable and predictable.
With 4 tools, the count is reasonable for a resource portal server, covering browsing (categories, glob), searching (grep), and reading (read). It is slightly lean but well-scoped, as each tool serves a distinct and necessary function without redundancy.
The tool set provides a complete workflow for discovering and accessing articles: listing categories, finding articles by path or content, and reading content. Minor gaps might include operations like creating or updating articles, but for a read-only portal, the coverage is sufficient for typical agent tasks.