Connects Claude to GitHub's GraphQL API to find genuinely beginner-friendly open source issues. The key trick here is that it checks for linked pull requests to filter out issues already claimed but not yet closed, something basic search misses. Returns AI-friendly summaries of issue bodies rather than dumping full text. Works unauthenticated but you'll hit rate limits fast. Better to pass a GitHub token via environment variable or use GitHub CLI auth. Targets repos with good-first-issue type labels but also attempts to spot beginner-appropriate work without explicit tags. Useful when you want to contribute to OSS but tired of finding every "available" issue already has someone working on it.
AI Scout for truly available, beginner-friendly OSS issues.
Run directly via npx:
npx mcp-oss-onramp
You can also run the server with a token directly in your terminal:
MCP_OSS_ONRAMP_GITHUB_TOKEN="your_token_here" npx mcp-oss-onramp
Add this to your AI agent's configuration:
{
"mcpServers": {
"oss-onramp": {
"command": "npx",
"args": ["-y", "mcp-oss-onramp"]
}
}
}
MCP_OSS_ONRAMP_GITHUB_TOKEN environment variable, ORgh auth login.Note: The server works without authentication but will be rate-limited and provide lower precision (no PR cross-referencing).
MIT
ray0907/git-mcp-server
cyanheads/git-mcp-server
io.github.b1ff/atlassian-dc-mcp-bitbucket
io.github.b1ff/atlassian-dc-mcp-jira
com.mcparmory/atlassian-jira
sirlordt/vscode-terminal-mcp