Conway agents, on-chain trust, preview-first execution.
Run wallet-backed agents with identity and reputation, publish TRUST.md on-chain, execute sandbox jobs, and expose MCP tools from one mission-control surface.
Wallet-backed agents | TRUST.md | on-chain reputation | sandbox jobs | MCP tools
Preview-first
Plan before every agent action
Idempotent retries
Safe replays on write routes
MCP-ready tools
API, jobs and docs in one surface
Agent Mission Panel
MCP tools available{
"identityRegistered": true,
"trustDocHash": "0x8d1d...9a42",
"jobsQueued": 2,
"warnings": ["high price impact on sell path"],
"blockingIssues": []
}
Plan (Preview)
Act (Execute)
Verify (Track)
Agent workflow
Plan, act and verify every operation with guardrails before write routes and on-chain execution.
Step 01
Plan (Preview)
Prepare agent actions with deterministic checks before any write route or on-chain call.
Step 02
Act (Execute)
Run guarded execution for agent registration, TRUST.md updates, jobs, launchpad or trade.
Step 03
Verify (Track)
Track async status, receipts, logs and artifacts with a single audit trail bundle.
Agent primitives
Core surfaces for Conway agent lifecycle, trust, compute, launchpad, and trading.
Identity & Reputation
Register agent identity on-chain and collect reputation attestations with deterministic writes.
TRUST.md Registry
Publish TRUST.md hash + URI and pin policy context to each agent lifecycle.
Sandbox Jobs
Run Linux jobs with command allowlists, plus downloadable logs and artifacts.
MCP Tool Server
Expose tools list and call bridge for agent-first integrations.
Launchpad
Create tokens with multi-agent fee split and preview-first execution controls.
Trade
Buy and sell through preview, explicit confirm, execute, and tracked settlement status.
Safety Band
Operational controls for Conway agent execution, on-chain writes, and job automation.
What it prevents: no execute without explicit confirm, unsafe params before chain, replay abuse on write routes, and weak webhook/rate-limit posture.