Discover
Index the official registry. Preserve a distinction between publisher statements and independent measurements.
Origin · version · transport Model Context Protocol / Evidence infrastructureAN OPEN DISCOVERY LAYER · EST. 2026
The registry tells you who's there. Eviduct measures what responds.
A server listing is not proof of reliability. We turn direct MCP observations into evidence your agents can actually use—without paid placements, invented ratings or guesses presented as fact.
DISCOVERY IS FREEINDEPENDENT OBSERVATIONSPRIVATE EVIDENCE AVAILABLE TO TEAMS
01 / THE METHOD
Registration isn't verification.
MCP directories describe software. They do not tell an agent whether a server was actually reachable, what it returned, or how certain that observation was.
Observe
Perform bounded, read-only MCP handshakes. Record response, latency, tool schema and observation time.
Timestamp · result · schemaDecide—or abstain
Compare qualified evidence using explicit sample thresholds. If there isn't enough data, don't invent a winner.
Provenance · confidence · limits
02 / THE EVIDENCE
An observation is more valuable than a star.
Review what our probes saw, not what somebody wrote about themselves. Repeated samples can show change and uncertainty—but a successful handshake never guarantees that a tool produces correct output.
Try the read-only discovery tools| Signal | What it establishes |
|---|---|
| Handshake | The endpoint responded to MCP initialization from our probe environment. |
| Tool inventory | Tools and input schemas advertised at observation time, without executing them. |
| Latency | Measured response time for our specific request from one region. |
| Change history | Whether repeatedly observed tool schemas differ between captures. |
| Confidence | Conservative uncertainty bound, based on actual sample size and time window. |
Nothing here proves payment settlement, tool correctness, or ownership. Those require separate verification and are never silently inferred.
03 / TRUST MODELPublic discovery.
Public discovery.
Private evidence.
Two surfaces, clearly separated. Useful enough to try without an account. Protected where the data and service economics matter.
01 / OPEN ACCESS
Discover.
Use our public MCP connection to search registry entries and read source-labeled, structured metadata.
No credentials · No provider executionConnect ↗
02 / PRIVATE ACCESS
Verify.
Independent measurement reports, evidence-qualified comparisons, per-company monitored watchlists and private usage statistics.
Scoped company API key
04 / INTEGRATE
Already using an agent?
All connection instructions
Already using an agent?
Give it a better source.
The free MCP integration is a real endpoint—not a screenshot. Add it to your AI client, inspect the discovered tools, and search structured data directly.
CODEX / TERMINAL
codex mcp add eviduct --url https://eviduct-mvp.onrender.com/mcp/public
FOR ENGINEERING TEAMS
Bring evidence to an actual infrastructure decision.
Tell us which agents and MCP servers your team depends on. We'll provision a scoped pilot key and evaluate the available signal coverage before agreeing on commercial terms.
PRIVATE PILOT / LIMITED ACCESS