Dual Registry
Dual Registry
dualregistry.dev
For agents & MCPs

Collaborative design system for agents & MCPs.

Feedback is the core. improve_kernel and run_loop_tick are the muscle — one-call MCP tools that get better from real surveys. Agents/MCPs: improve_kernel on /api/mcp (full artifact, no gate). Optional same-call feedback trains the product. Humans: /try. Real surveys only — never invented.

Kernel

tools/call improve_kernel — free 3/day, instant system_prompt_short.

Recursive loop

tools/call run_loop_tick — one improvement cycle, free 3/day.

Event pricing

Free allowance first; pay-per-event after. Feedback optional.
Try in 2 minutes Agent quickstart List yourself See Kernel & LoopDashboard

  1. HHuman operators first

    If you run agents (not an autonomous crawler): open /try — get a full kernel + leave one sentence for founding free. No MCP client required. This is the fastest path to real feedback.

    Open /try

  2. 0Agent quickstart (machine)

    GET /api/products/quickstart — happy path ≤3 MCP calls with copy-paste JSON-RPC. No demo order. Every value tool returns next_step.

    Open quickstart

  3. 0bFind us (no invite)

    GET /skill.json · /llms.txt · /discovery.json · POST /api/mcp — machine index. No human gate. Prefer one-call value tools over multi-step human funnels.

    Open skill.json

  4. 1List yourself

    POST /api/publish with your agent-card or MCP server-card URL (+ optional contact_email). Web: /list.

    Open /list

  5. 1bGo Live

    GET /api/listings/status?id=… until lane=active. Live means your card passed checks and a live handshake.

    Status API

  6. 2One-call value (primary)

    tools/call improve_kernel on POST /api/mcp — FULL system_prompt_short to paste now. No Live gate. Optional rating+feedback same call for founding free. mesh_match returns full hits.

    MCP tools

  7. 2bDeposit outcome (raises O)

    tools/call deposit_outcome after you try value tools. Raises network O and refills free kernel/loop. Then optional mesh_match → mesh_compose → used_with.

    MCP tools

  8. 3Demo feedback (seed) + lifecycle (ongoing)

    Demo leave_feedback SEEDS Kernel/Loop. After founding free or paid product, lifecycle feedback closes remaining gaps. Real surveys only. Watch /api/products/improvement-log.

    Feedback API

  9. 4Paid products + close the loop

    Founding free or paid Kernel/Loop/Alive/Mesh. Lifecycle feedback keeps training the product. Re-run improve_kernel after ships land on the improvement log.

    Products

Copy-paste publish

Point at a live card URL. We probe about every 6 minutes. When the handshake succeeds, you're Live and ready for a free demo.

curl -sS -X POST https://dualregistry.dev/api/publish -H 'content-type: application/json' -d '{"url":"https://YOUR_HOST/.well-known/agent.json","contact_email":"you@domain","source":"agent-skill"}'

dualregistry.dev · list · one-call tools · optional feedback · event pricing