Survivor · Gate Gate Live
Execution Governor · MCP-Native · Chain-Agnostic

Execution without judgment
is a vulnerability.

Agents can reason and plan. The hard part is controlling when they are allowed to act. SURVIVOR enforces programmable constraints at action-time — evaluating live token risk, agent reputation, and identity signals before any irreversible operation reaches the chain.

Live Gate Health ↗ GitHub ↗
Agent Intent
→
MCP Tool
→
SURVIVOR Gate
→
Execution Target
ALLOW THROTTLE READ_ONLY DENY
What It Does
ALLOW Risk acceptable. Full execution forwarded with original intent intact.
THROTTLE Elevated risk. Execution continues with constrained notional, slippage, and cooldown.
READ_ONLY Risk too high for live execution. Simulation and quote-only paths permitted.
DENY Hard stop. Structured reason codes returned for audit. No capital at risk.
Live Demo
POST /gate · production 
curl -s -X POST https://survivor-oracle-production-1501.up.railway.app/gate \
  -H "content-type: application/json" \
  -d '{"chain":"solana","from_asset":"SOL","to_asset":"<MINT>",
     "notional_usd":2500,"slippage_bps":100,"kind":"swap"}'
{
  "decision": {
    "ok": true,
    "mode": "SIMULATE",
    "policy": {
      "decision":   "READ_ONLY",
      "confidence": 0.70,
      "reasons": [
        { "code": "LP_UNLOCKED",          "severity": "high" },
        { "code": "NO_LIQUIDITY",         "severity": "high" },
        { "code": "HOLDER_CONCENTRATION", "severity": "high" }
      ],
      "expires_at": 1771890169
    }
  }
}
Behavior-Aware Execution

SURVIVOR evaluates not just the token — but the agent requesting execution. AIS (Agent Integrity Score) is computed from on-chain behavioral data across multiple protocols. Tier directly gates execution privilege.

AIS Tier
Execution Privilege
Example Limit
TRUSTED
Full execution, highest limits
$5,000 / tx
STANDARD
Normal limits, standard policy
$2,000 / tx
FLAGGED
Reduced position, tighter slippage
$200 / tx
RESTRICTED
Execution denied regardless of token score
$0

A token scoring 62 (CHALLENGE) routes differently depending on the caller. A TRUSTED agent may proceed at $2,000. A FLAGGED agent is capped at $200 or denied. Same token. Different judgment.

Judgment Signals · v1
Where SURVIVOR Fits
On-chain Behavior
What did the agent do across protocols?
VERITY AIS
How trustworthy is the agent's track record?
IAM Engine
Who is the real actor behind the wallet cluster?
SURVIVOR Gate
Should this action be allowed right now?
Signed Attestation
Auditable proof of every decision made.

MCP never knows SURVIVOR exists. SURVIVOR never executes trades. This preserves neutrality — invisible infrastructure that every agent stack routes through.

Request Access

Evaluating SURVIVOR Gate for agent stacks or high-impact automation?

DM @youngs_modulus  ·  contact@identityaware.ai

Contact ↗