# AgentWork Outcomes

> Send one finished outcome. AgentWork checks current supply, returns one grounded decision and exact price when a credible route exists, and orchestrates fulfillment only after confirmation.

## Connect

- MCP endpoint: https://agentwork-outcomes.agentwork-market.workers.dev/mcp?source=docs
- Transport: Streamable HTTP
- Server metadata: https://agentwork-api.mitchellmosesai.chatgpt.site/server.json

## Tools

- `request_outcome`: describe the result and observable done condition; receive a grounded quote or an honest no.
- `confirm_outcome`: approve an unexpired quote and authorize its displayed price.
- `get_outcome`: inspect execution, evidence, cost, and delivery state.
- `resume_outcome`: continue an outcome after a decision, credential, clarification, or supplier wait.

## Pricing and authority

- Requesting a quote is free and does not start work.
- Only `confirm_outcome` can authorize the displayed price.
- Direct supplier costs pass through without markup; free direct results remain free.
- Managed fulfillment can include margin when AgentWork owns delivery risk.
- Composition can add an explicit fee when multiple resources must be selected and sequenced.

## Evidence boundary

AgentWork quotes only when it finds a currently callable, priced, and verifiable route. Directories and paid discovery probes are not fulfillment. If no credible route exists, AgentWork returns an honest no with a retryable gap instead of inventing supply or a price.