MCP Server
Connect any MCP client to PT-Edge for live AI ecosystem intelligence. 8 tools, 220K+ repos, real-time data.
Connection
Streamable HTTP (Claude Desktop, SDK clients)
https://mcp.phasetransitions.ai/mcp/stream
JSON-RPC POST (Claude.ai web connector)
https://mcp.phasetransitions.ai/mcp
Authentication
Pass your token as a query parameter or Authorization header:
?token=your_token Authorization: Bearer your_token
Both the legacy API token and pte_* API keys are accepted. Get a free API key for authenticated access.
Available Tools (8)
Slim tool surface following the Oakbridge pattern: discover the schema, compose SQL, get answers.
get_status()
Start here. Returns table count, repo count, domains, freshness, and guidance.
list_tables()
All database tables with row counts. Use before querying.
describe_table(table_name)
Column names, types, and nullability for a specific table.
search_tables(keyword)
Find tables by keyword in table or column names.
query(sql)
Run any SELECT query. Read-only, 5s timeout, 1000 row limit. The workhorse.
list_workflows()
Pre-built SQL recipe templates for common questions. Adapt and run via query().
find_ai_tool(query, domain?, limit?)
Semantic + keyword search across 220K+ AI repos. Describe what you need in plain English.
submit_feedback(topic, correction, category?)
Report bugs, request features, or share observations about the data.
Recommended Workflow
get_status()— see what's availablelist_tables()— browse the schemadescribe_table('ai_repos')— see columnslist_workflows()— get recipe templatesquery('SELECT ...')— get your answer
For semantic similarity search, use find_ai_tool() instead of SQL.