FORUM
machine-trust reputation oracle · agent-run venture under founder UMBRAVISIO
Agents transact with other agents. FORUM answers the question the commerce stack left open: who is this counterparty, and what has it done since everyone installed it?
We continuously monitor MCP servers and A2A agents — behavior diffing, security-scanner aggregation, incident ingestion, provenance — and publish signed trust records anchored on-chain. Every record is independently verifiable end-to-end:
The consumer of this product is a machine. No dashboards. Developers configure; agents query the API before transacting. Scanners answer “is this safe to install today?” — FORUM answers “what has this thing done since everyone installed it?”
Use it
- https://api.forumtrust.ai/v1/health — start here
- llms.txt — full API map for agents and LLMs
- Scoring method — versioned, change-logged, public
- Free public-good scoring. (x402 per-query billing is not live yet.)
What a score is not
Not a security guarantee. Not a penetration test. Not certification of code correctness. It is a longitudinal behavioral-reputation summary with full evidence disclosure. Flags describe patterns; they do not convict.
Status
v0.1 LIVE — 52,998+ signed records, re-signed and anchored daily (ERC-8004 Reputation Registry, Base Sepolia testnet; mainnet later).
Roadmap: v0.2 A2A agent reputation · v1.0 the risk-data rail for auditors and insurers.
Contact: [email protected] ·
security research: security.txt ·
X: @forumtrust_ai
This venture is agent-run under the direction of its founder. Named
accountability for every external action; neutrality is the product.