SeenRelay
Server Details
Avoid redundant expensive validation. CHECK recent evidence; OBSERVE fresh independent results.
- Status
- Healthy
- Uptime
- 100.0% over the last 24 hours
- Last Tested
- Transport
- Streamable HTTP · MCP 2025-11-25
- URL
- Repository
- ovladon/seenrelay
- GitHub Stars
- 0
- Server Listing
- SeenRelay
- ChatGPT
- –
- Claude
- –
TDQS
Scored across 2 tools
The two tools are cleanly split into a read/check operation and a write/observe operation with explicitly opposite timing: CHECK before repeating a read, OBSERVE after a fresh validation. The descriptions reinforce this boundary by stating what each tool does not do, leaving little chance of misselection.
Both names follow the same verb_noun convention with a shared object noun: check_fact and observe_fact. The pattern is fully predictable and readable.
Two tools is thin relative to the typical 3–15 range, but for a narrowly scoped observation relay the read/write pair is coherent and each tool clearly earns its place. The surface is minimal by design rather than accidentally sparse.
The core lifecycle is present—observe a result, then check it later—but there is no way to list, inspect, revoke, or manage stored observations, leaving agents with limited visibility beyond a single CHECK. Some of this may be intentional immutability, but it is still a notable gap for fleet-level use.
Available Tools
2 toolscheck_factCheck Fact FreshnessARead-onlyIdempotentInspect
Avoid a duplicate expensive read-only tool call before repeating source-backed validation, especially across an agent fleet. Use for repeated paid web search, metered scraping, browser/extraction, rate-limited APIs, duplicate API validation or multi-step validation when the caller already has a known value. Returns SAME_OBSERVED, CHANGED_OBSERVED, CONTESTED, STALE or UNKNOWN. Only SAME_OBSERVED can be considered for caller-policy-gated reuse; every other status requires normal authoritative validation before OBSERVE. SeenRelay never browses or verifies externally.
Inputs: fact (required), known_value (required), max_age_seconds. Returns: CHECK result. Exactly one of a decision status or an error is expected; additio…
observe_factContribute Fact ObservationAInspect
After independently performing a fresh paid search, scrape, browser extraction, API read or other source-backed validation, deposit the observed result for later CHECKs. Returns accepted=true when a new observation is stored or deduplicated=true for a recognized retry/recent duplicate. Never OBSERVE hearsay. Optional Ed25519 proof establishes key possession and continuity, not truth or real-world independence.
Inputs: fact (required), value (required), observed_at, observer_id, +4 more. Returns: OBSERVE result. accepted=true means a new observation was stored; deduplicated=…
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Related MCP Connectors
Physical-world evidence and operability checks with provenance and explicit data gaps.
Second opinion before an irreversible agent action; signed proofs, free verify, public ledger.
Verify a company, figure or citation before you rely on it. Card or crypto; signed receipt.
Verify stock signals, replay evidence, provenance, freshness, and caveats.
Related MCP Servers
- AlicenseNot gradedqualityBmaintenanceEnables autonomous AI agents and MCP clients to validate observation records so unverified states can pass while self-declared VERIFIED claims are blocked unless an independent external verifier is present.MIT
- AlicenseCqualityCmaintenanceEnables agents to validate physical-AI evidence through synthetic scenario and episode reviews, explain beliefs from delayed or conflicting evidence, and audit simulation replays while tracing why a workflow pauses. Pairs local freshness checks with bounded cloud model review and exposes no actuator or arbitrary filesystem tools.5Apache 2.0
- AlicenseNot gradedqualityCmaintenanceEnables checking what a device or agent reported against independent evidence, answering Worked, Problem found, or Not verified while always showing the evidence and stating its limits. Adds a night check that aggregates results across devices and keeps timing straight, leaving items without independent evidence as Not verified instead of guessing.Apache 2.0
- AlicenseBqualityCmaintenanceEnables AI agents to verify technical claims against supplied evidence, identify unsupported assumptions and contradictions, and recommend the smallest next check before acting.5MIT
Glama MCP Gateway
Add one secure layer between your agents and this server.