> ## Documentation Index
> Fetch the complete documentation index at: https://docs.epode.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Real-world integration patterns

> Choose outcome boundaries and session references for search, crawl, browser, docs, and authenticated operations products.

The best configuration follows the product interaction the customer agent already performs. Do not invent
an agent identity or an Epode-only session.

| Product pattern               | Observe                       | Ask for feedback                                   | Session reference                   |
| ----------------------------- | ----------------------------- | -------------------------------------------------- | ----------------------------------- |
| Search or retrieval API       | Selected result responses     | Each agent-requested result                        | Existing agent run ID, if available |
| Async crawl or export         | Terminal result               | Completed or failed outcome the agent can evaluate | Existing job or run ID              |
| Browser automation MCP        | All browser tools             | Close/finalize, or another real journey boundary   | Browser session ID                  |
| Documentation MCP             | Resolve and query tools       | Answer-bearing query                               | Agent run ID                        |
| Email, payments, or issue MCP | Selected side-effecting tools | Each completed business operation                  | Authenticated workflow ID           |

## Search APIs with shared caching

If ordinary traffic shares a CDN response, use `cacheMode: "request"`. A feedback-aware caller sends
`Agent-Feedback-Request: 1`; public callers keep the public cache policy.

[Run the search example](https://github.com/open-software-network/os-epode/tree/main/examples/icp-search-express)

## Async APIs

Use `shouldInstrument` to expose feedback only when the body reaches a terminal status. This avoids asking
the agent to evaluate a job it has not used yet.

[Run the crawl example](https://github.com/open-software-network/os-epode/tree/main/examples/icp-crawl-fastify)

## Stateful MCP without micro-feedback

Use `includeTools` for the complete trace and `feedbackTools` for the final boundary. A result-aware
`sessionRef` extractor groups the session-creation result with later calls.

[Run the browser example](https://github.com/open-software-network/os-epode/tree/main/examples/icp-browser-mcp)

## Authenticated, sensitive operations

Derive `customerRef` from verified API or OAuth context, never an email address or a tool argument. Epode
rejects feedback containing likely personal data, secrets, prompts, transcripts, or raw product content.
The agent may retry the same handle with a safe abstract summary.

[Run the authenticated operations example](https://github.com/open-software-network/os-epode/tree/main/examples/icp-operations-mcp)