# AsyncWork

> AsyncWork is a callable network of human-curated, expert-maintained Small Language
> Models (SLMs). Named human experts each maintain a model of their craft; an agent
> calls AsyncWork, which routes to the right expert SLM(s), synthesizes one answer, and
> returns it with provenance (which experts, which sources, a confidence signal). For
> high-stakes work, the agent can escalate the same thread to the named human expert.
> AsyncWork operates three properties: a commercial advisory (asyncwork.com), a
> community/social-impact arm (asyncwork.org), and this agent-first interface (asyncwork.ai).

## What an agent buys (two tiers)
- Synthesized SLM answer — one answer from one or more expert SLMs, with provenance + confidence. Machine-answered, per-call micropayment. (planned, v3)
- Human verification / advisory — a named expert reviews, corrects, signs, or advises. Per-engagement, revenue-shared with the expert. (planned, v4)

## Properties
- [AsyncWork.com — human program delivery](https://asyncwork.com/llms.txt): niche-domain Asyncwork Advisors deliver expert guidance to clients.
- [AsyncWork.org — community arm](https://asyncwork.org/llms.txt): human mentors + apprentices co-deliver community programs (e.g. FamilyCyberSecurity).
- [AsyncWork.ai — this site](https://asyncwork.ai/): the expert-SLM network for agents.

## For agents
- [Agent onboarding guide](https://asyncwork.ai/agents.md): register, identify, call, pay, escalate.
- [A2A Agent Card](https://asyncwork.ai/.well-known/agent.json): capabilities, endpoints, auth schemes, skills.
- [MCP manifest](https://asyncwork.ai/.well-known/mcp/manifest.json): callable tools for MCP clients.
- [Service catalog](https://asyncwork.ai/catalog.json): properties, the two-tier product, services, commerce terms.

## Services
- get_catalog / search_knowledge — discovery + public knowledge (anonymous, free).
- ask_expert — synthesized expert-SLM answer with provenance (registered or pay-per-call via x402; planned).
- verify_with_human — escalate to the named human expert (registered + AP2 mandate; planned).

## Commerce (rail-neutral)
- Default per-call: x402 (USDC), native to AsyncWork's Cloudflare stack (planned).
- Trust layer: AP2-style verifiable mandates for premium spend (planned).
- Provenance: every answer is sourced; premium answers carry a signed verification credential.

## Optional
- [Full machine map](https://asyncwork.ai/llms-full.txt)