Tenders, registries, permits and filings are public — but they sit across thousands of sites in dozens of formats that no agent can use. MCPie is an open network that collects them, cross-checks them, and serves them as clean, traceable, payable verticals.
The supply side of the agent economy doesn't exist yet. Models are capable, but the real-world data they need is fragmented across national portals, municipal sites and PDF archives. Turning that into clean, payable data is the single biggest gap between AI agents and the real world.
Incumbents solve it with 100+ human researchers per vertical and sell the result for $15–30K a seat. That moat is exactly what a contributor network inverts.
Everything runs on one repeating cycle. Contributors run the open-source client; the network validates, publishes, and keeps the data honest as it drifts.
Run the MCPie client to pull public records into the network — using the AI subscription you already pay for.
The client cross-checks each record against other contributors' before it's accepted. Disagreements are rejected, not averaged.
Validated records go live as signed verticals that agents can query immediately.
The client re-checks published data as it drifts — feeding straight back into 01, so the set never goes stale.
Every record carries evidence of where and when it was collected, so a field can always be traced back to its source document.
A record is only published once independent contributors agree on it. Conflicts are surfaced and rejected rather than silently merged.
Published data is re-checked continuously. When a source changes, the record is rebuilt instead of quietly rotting.
MCPie serves tender, registry and permit data through MCP or a normal API — every field traced to its source, and you pay only for what you call. Each vertical ships as an MCP server with an x402 endpoint, so agents pay per query in USDC.
The first verticals are forming now. Which verticals grow next is decided by where contributors and demand actually point.
MCPie