Find the right crypto stack with evidence-backed protocol intelligence.

How it works

From maintainer evidence to agent-ready answers.

Composewith turns reviewed protocol evidence into structured outputs that agents can query, cite, and use when recommending a stack. Credible neutrality is the constraint: recommendations are based on public evidence and release proofs, not paid placement.

How to read protocol cards

Readiness is confidence. Difficulty is effort.

The number on each card is a 0-100 readiness score computed from public evidence. The difficulty pill is separate: it estimates integration effort. A protocol can be high-readiness and still high-difficulty if it is well documented and production-proven but requires complex contracts, indexing, payments, or operational controls.

Score breakdown: 30 points for integration surfaces, 40 for capability depth and production confidence, and 30 for chain coverage, freshness, and critical issue checks.

Agent-facing output

What the agent receives

The response is structured so an agent can cite it, turn it into a build plan, or hand it to a developer without guessing which protocol fills which layer.

{
  "intent": "agent-native services marketplace on Arbitrum",
  "matched_intent": "build-agentic-commerce-with-x402",
  "chain": "arbitrum-one",
  "recommended_stack": [
    {
      "slug": "x402",
      "role": "payments_and_commerce",
      "reason": "Use x402 for HTTP 402 payment-required flows so buyer agents can pay per request with bounded spending policies."
    },
    {
      "slug": "erc-8004",
      "role": "agent_identity_and_trust",
      "reason": "Use ERC-8004 and 8004scan for public agent identity, discovery, reputation, and validation records."
    },
    {
      "slug": "arbitrum-one",
      "role": "settlement_or_execution",
      "reason": "Use Arbitrum One for low-cost EVM marketplace contracts, payment receipts, refund state, and operational controls."
    },
    {
      "slug": "goldsky",
      "role": "indexing_and_data",
      "reason": "Index listings, x402 payment events, fulfillment status, feedback, and agent reputation surfaces."
    },
    {
      "slug": "safe",
      "role": "supporting_protocol",
      "reason": "Manage treasury, refunds, facilitator configuration, and incident approvals with multisig controls."
    }
  ],
  "excluded_candidates": [
    {
      "slug": "uniswap",
      "reason": "Useful for liquidity and swaps, but not the primary payment or agent-trust layer for this marketplace."
    }
  ],
  "risks": [
    "ERC-8004 is draft status; track spec and contract updates before production.",
    "x402 handles payments, not fulfillment, escrow, returns, or dispute resolution.",
    "Buyer agents need spending caps, replay-safe order IDs, receipt validation, and refund paths.",
    "Reputation needs Sybil resistance and reviewer trust rules, not just raw feedback counts.",
    "Indexer lag can make listings, payment state, or fulfillment status stale."
  ],
  "next_steps": [
    "Define listing, order, fulfillment, refund, and dispute state separately from x402 payment state.",
    "Register seller agents through ERC-8004 and publish MCP/A2A endpoints in their registration files.",
    "Verify x402 facilitator, token, and network support for Arbitrum before production.",
    "Add event indexing for listings, payments, fulfillment, feedback, and agent status.",
    "Write failure drills for failed payment, stale indexer data, bad fulfillment, and malicious feedback."
  ]
}

Arbitrum architecture sketch

From discovery to paid fulfillment

• Discovery: seller agents register through ERC-8004, expose MCP/A2A endpoints, and become browsable through 8004scan.
• Payment: paid service endpoints return HTTP 402; buyer agents retry with x402 payment headers and bounded spending policies.
• Settlement: Arbitrum One handles marketplace contracts, payment receipts, refund state, and operational controls.
• Indexing: Goldsky or The Graph indexes listings, payment events, fulfillment status, and feedback for agent-readable retrieval.
• Controls: Safe manages treasury, refunds, facilitator configuration, and incident approvals.

Main risks to verify: ERC-8004 draft status, x402 facilitator and asset support on Arbitrum, payment replay/idempotency, agent spending limits, stale listing data, failed fulfillment, refund policy, and Sybil-resistant reputation.