data-ai MCP Server
Real-time trustworthiness evaluation and safety interception for AI agents. Semantic analysis, safe alternative suggestions, multi-step attack chain detection, and LLM-as-Judge.
Discovered via github-topic:mcp and last synced 2mo ago.
Install instructions not detected yet
Check the source repository for the latest setup steps.
MEDIUM
Malicious dependency injection
Eval injection, subprocess spawning, remote code execution
No
BLOCK
API keys in logs, tokens in URLs, secrets in world-readable files
**95.0%**
What it does
Modifying SSH config, disabling firewalls, altering PATH
AgentTrust
`curl -o script.sh url && cat script.sh && bash script.sh`
Status
Yes
Steps
Examples
Verdict Accuracy (production default)
No
Intentional data wiping
Accidental deletion, overwriting config files, writing to system paths
**100%**
170 policy rules (incl. catastrophic-drop & protected-branch force-push BLOCK rules), Shell normalizer (11 strategies, incl. long-flag canonicalization + base64 decode), chain-alert verdict escalation, RiskChain parameter-aware matching, opt-in LLM-Judge fallback, safety contracts (13 tests), capability-control (least-privilege) mode, 630-scenario held-out verdict testing (95.4% production default; no overfitting gap vs the internal 300)
Requests to internal IPs, unencrypted data transmission, DNS exfil
**92.0%**
Yes
Piping sensitive files to external endpoints, steganography
Operator promotion review UI + feedback-driven rollback, Shell AST parsing, richer Judge cache strategies, plugin ecosystem
https://docs.agenttrust.tech
`CapabilityNamespace`, `AttestorProfile`, `ValidationRequest`, `ValidationAttestation`
`cd trustgate/server && pnpm test`
Detail
`multisig_threshold_enforced` — distinct signer count ≥ threshold
`gate_payment_strict_correctness` — strict Ok ⇔ Allow + 3 disjoint variants
`cargo test --workspace --lib`
`counterparty_tier_monotone` — strict pass ⇒ loose pass
`npm i @agenttrust-sdk/trustgate`
Sub-checks
`anchor test --provider.cluster devnet`
`paused_implies_no_allow` — KillSwitch paused ⇒ never Allow
`register_namespace`, `register_attestor`, `request_validation`, `respond_to_validation`, `revoke_validation`
`velocity_counter_le_limit` — Allow preserves cumulative ≤ max
tx
`validation_expiry_correct` — expired attestation ⇒ never Allow
Count
Sandbox (E2B / gVisor / seccomp)