Endpoint · inferencePOSTcanaryguard-402.onrender.com /v1/security/mcp-tool-poisoning-audit
Inspects Model Context Protocol (MCP) tool definitions, descriptions, and JSON schemas for hidden prompt overrides, shadowing attacks, and exfiltration sinks.
Overview
Grade?
UNRATED
0 of 30 paid calls toward a letter
Price
$0.02 per call
Paid calls
none yet
delivery unknown until someone pays
More info
Seller?
URL
https://canaryguard-402.onrender.com/v1/security/mcp-tool-poisoning-audit
MCP server on this host: io.github.ak10082247-max/canaryguard-402
Seen
first 32 days ago · last 26 days ago
17 free handshakes
Payment
Pays through
x402 · MPP evm
Offered on?
Base USDC, Solana
Listed on?
Base
Score, from free handshakes onlyweights 2026.09.6
UN
RATED
RATED
0.534 liveness verified, no letter
To a letter · 30 paid calls
0 of 30
Held back by sample size, not by the seller · gap to 1.000 is 0.466: 0.237 uncertain ? + 0.229 short ? · 0.116 to C
| Component | Weight | Measured | Lower bound ? | Adds | Short ? | Uncertain ? | |
|---|---|---|---|---|---|---|---|
| livenesscosts the most it answered at all |
0.60 | 0.95 | 0.557 | 0.334 | −0.029 | −0.237 | |
| latency p95 it answered as fast as its class |
0.20 | 0.00 | 0.000 | 0.000 | −0.200 | 0 | |
| price stability the price stayed where it was listed |
0.20 | 1.00 | 1.000 | 0.200 | 0 | 0 | |
| Composite | 1.00 | 0.534 | gap to 1.000 = 0.466 · 0.229 short · 0.237 uncertain | 0.534 | −0.229 | −0.237 |
Not measured: correctness · nobody paid; honesty · nobody paid; schema conformance · no answer had a published shape to check
The multiplication, written out
0.600 × 0.557 liveness
+ 0.200 × 0.000 latency p95
+ 0.200 × 1.000 price stability
= 0.534 → UNRATED: 17 of 30 paid calls weights 2026.09.6 · 17 samples · seed cap_01M546ENXVA3R02KM97M2KEFV9|2026-10-07T09:05:50.648Z|2026.09.6
Verify it yourself
npx teppi-check https://canaryguard-402.onrender.com/v1/security/mcp-tool-poisoning-auditcurl -s https://api.teppi.xyz/v1/trust/cap_01M546ENXVA3R02KM97M2KEFV9