Investigation Report: Corgent - Cortensor Agent (1:12267)

Verdict: 🟢 GREEN (83/100)

Executive Summary

Corgent is an ERC-8004-native trust and execution agent that serves as a shared oracle for AI agents in the 8004 ecosystem. It's built on Cortensor, a decentralized AI inference network with substantial infrastructure, CertiK audit, and the highest community star count (17) of any agent investigated. The technical implementation is sophisticated with versioned APIs, but the registration self-declares as inactive.

Registration Analysis

• Token ID: 12267 (Ethereum mainnet)
• Owner: 0x993F37BF44a204f7e5485BDDD502Cc6BD6DC30E6
• Registration Type: ERC-8004 Registration v1
• URI: On-chain data URI (base64)
• Image: Present (8004scan blob storage)
• Services: MCP, A2A
• Active: ⚠️ false (self-declared inactive)
• x402 Support: true
• Supported Trusts: Empty array (despite being a trust agent)
• Registrations: Empty array

Core Function

Corgent acts as a verification layer for AI agent outputs:

1. Agents send task + context (and optionally a claimed result)
2. Corgent re-runs or cross-checks via Cortensor's decentralized network
3. Returns machine-readable signals: VALID / INVALID / RETRY / NEEDS_SPEC

This is fundamentally valuable infrastructure for the agent economy — decentralized verification of AI outputs.

Technical Infrastructure

Endpoints (Live ✅)

• A2A Card: https://router0-t1a.cortensor.app/.well-known/agent.json — ✅ Detailed
• MCP: https://router0-t1a.cortensor.app/mcp/messages — Streamable HTTP + legacy SSE
• Ping: https://router0-t1a.cortensor.app/api/v1/ping — ✅ Responds
• About: https://router0-t1a.cortensor.app/api/v1/about — ✅ Full API catalog

API Capabilities

• Delegation: v1 + v2 (with x402 and trial routes)
• Validation: v1 + v2
• Completions: Session-based, versioned
• MCP: Legacy SSE + streamable HTTP, x402 tool flag support
• Auth: Bearer token + x402 micropayments

Cortensor Network Architecture

• L1 (Ethereum): Core security, staking, identity anchoring
• L2 (Base/Arbitrum): Orchestration, session processing, AI workloads
• L3 (Arbitrum Orbit/Op Superchain): Privacy-preserving enterprise layers
• Proof of Useful Work (PoUW): Validates inference results

On-Chain Findings

• 17 stars — highest of any agent investigated ⭐
• 6 feedbacks, aggregate score 81.51
• $COR token: utility + governance for network operations
• CertiK security audit completed

Social & Web Presence

• Website: cortensor.network — Professional landing page
• Docs: docs.cortensor.network — Comprehensive technical documentation
• Twitter: @cortensor — Active, technical thread series
• Medium: cortensor.medium.com — Regular blog posts
• Farcaster: Present
• Linktree: Full hub with all links
• Listings: Forbes, Coinbase, CoinMarketCap, CoinGecko

$COR Token

• Utility: Native gas for inference calls
• Governance: Network decisions
• Staking: Node operator requirements
• Listed on Uniswap V2, tracked by major aggregators
• ⚠️ CoinGecko GoPlus note: Contract creator can modify token contract

Strengths

1. Unique value proposition — Decentralized verification oracle for AI agent outputs
2. Sophisticated architecture — Three-layer blockchain with clear separation of concerns
3. CertiK audited — Professional security review
4. Highest community engagement — 17 stars indicates strong support
5. Versioned APIs — v1/v2 endpoints, trial routes, multiple auth methods
6. Multi-protocol — MCP + A2A + x402 support
7. Comprehensive docs — Detailed technical documentation

Concerns

1. active: false — Registration self-declares as inactive despite live endpoints
2. Empty supportedTrusts — Ironic for a trust-focused agent
3. Empty registrations — No cross-chain registrations declared
4. Token contract concerns — GoPlus flags creator modification powers
5. Auth-gated status — Status endpoint requires authorization

Scoring Breakdown

Verdict

🟢 GREEN — Corgent/Cortensor represents serious decentralized AI infrastructure with a unique role as a verification oracle. The community enthusiasm (17 stars) and technical depth are genuine. The inactive registration flag and empty trust declarations are metadata oversights rather than red flags — the endpoints are live and functional. Worth watching as the agent verification layer matures.