# MandAPI Cookbook (en)

Base API: https://api.mandapi.com/v1
Examples require an account API key and a model available to that key.

- [Build a small streaming chat app](https://mandapi.com/en/cookbook/get-started/chat-app.md): List your available models, send one message, stream the reply, and keep a local conversation using the OpenAI-compatible MandAPI endpoint.
- [Move an OpenAI-compatible chat integration to MandAPI](https://mandapi.com/en/cookbook/get-started/migrate.md): Change the chat connection boundary while preserving the application's existing messages, response handling and interface.
- [Connect Claude Code to MandAPI](https://mandapi.com/en/cookbook/coding-agents/claude-code.md): Configure the Anthropic-compatible endpoint, choose a supported model and check your first coding session.
- [Configure Claude Desktop with a MandAPI gateway](https://mandapi.com/en/cookbook/coding-agents/claude-desktop.md): Use Claude Desktop's third-party inference panel with an Anthropic-compatible MandAPI connection.
- [Review a selected code diff with MandAPI](https://mandapi.com/en/cookbook/coding-agents/code-review.md): Preview a small staged diff, then request a second opinion without changing or committing code.
- [Use Codex CLI with MandAPI](https://mandapi.com/en/cookbook/coding-agents/codex-cli.md): Add a custom Responses provider with an environment-based key, while keeping existing project settings.
- [Start a coding session in MandAPI Connect](https://mandapi.com/en/cookbook/coding-agents/connect.md): Use the Windows desktop app to open Claude Code or Codex CLI with your MandAPI account.
- [Use MandAPI for chat in Cursor](https://mandapi.com/en/cookbook/coding-agents/cursor.md): Set the OpenAI-compatible endpoint in Cursor and start with a supported chat model.
- [Connect Hermes Agent to MandAPI](https://mandapi.com/en/cookbook/coding-agents/hermes.md): Add MandAPI as a custom Chat Completions endpoint in Hermes.
- [Create a MandAPI model profile for Junie CLI](https://mandapi.com/en/cookbook/coding-agents/junie.md): Use Junie's custom JSON profiles with a complete Chat Completions request URL.
- [Use an MCP tool with a MandAPI-backed coding session](https://mandapi.com/en/cookbook/coding-agents/mcp-servers.md): Keep model inference in MandAPI while Claude Code connects to an MCP server locally.
- [Add MandAPI as an OpenClaw model provider](https://mandapi.com/en/cookbook/coding-agents/openclaw.md): Register a custom provider, select a real model and check a local chat before connecting channels.
- [Add MandAPI to OpenCode v1](https://mandapi.com/en/cookbook/coding-agents/opencode.md): Configure an OpenAI-compatible provider and select your MandAPI model in OpenCode v1.
- [Build an FAQ support chat in n8n](https://mandapi.com/en/cookbook/integrations/n8n-support.md): Download a small workflow, connect a MandAPI credential, and answer support questions from an editable business FAQ.
- [Describe objects in an image with MandAPI](https://mandapi.com/en/cookbook/integrations/vision.md): Send a local image to a current Vision-capable model through Chat Completions.
- [Build a small terminal interface for your agent](https://mandapi.com/en/cookbook/building-agents/agent-tui.md): Expose progress, reference reads, and resumable steps with a portable Python terminal interface.
- [Summarize an older conversation while keeping recent turns](https://mandapi.com/en/cookbook/building-agents/compact-context.md): Compact application-managed message history without promoting a summary to an instruction.
- [Draft with one model and review with another](https://mandapi.com/en/cookbook/building-agents/delegate.md): Coordinate two ordinary chat requests in your own application.
- [Stop repeated tool calls and runaway agent loops](https://mandapi.com/en/cookbook/building-agents/detect-loops.md): Enforce a step limit and detect identical calls using application trace data.
- [Build a headless agent with read-only tools](https://mandapi.com/en/cookbook/building-agents/headless-agent.md): Create a Python worker with local action validation, JSON events, saved state, and bounded chat requests.
- [Require human approval before an action](https://mandapi.com/en/cookbook/building-agents/human-approval.md): Keep a model's proposed action separate from permission to execute it.
- [Resume longer tasks from local checkpoints](https://mandapi.com/en/cookbook/building-agents/long-horizon.md): Keep progress across runs without resetting attempt limits or treating saved text as system instructions.
- [Classify supplied support messages into a fixed set of labels](https://mandapi.com/en/cookbook/evaluate-and-optimize/classify-text.md): Ask a chat model for JSON, then validate the result in your application.
- [Prepare reviewed examples for a smaller-model experiment](https://mandapi.com/en/cookbook/evaluate-and-optimize/distillation.md): Build a local JSONL dataset and holdout set without assuming a MandAPI training service.
- [Extract and validate fields from invoice text](https://mandapi.com/en/cookbook/evaluate-and-optimize/extract-invoice.md): Turn supplied invoice text into a small validated record using ordinary chat output.
- [Build multilingual RAG with local retrieval and MandAPI](https://mandapi.com/en/cookbook/evaluate-and-optimize/rag.md): Compute neural embeddings and reranking locally, then let MandAPI answer from the selected passages.
- [Check the boundaries of your own FAQ application](https://mandapi.com/en/cookbook/evaluate-and-optimize/red-teaming.md): Run harmless local cases for unknown questions, policy overrides and fabricated actions.
- [Export your application's usage events as CSV](https://mandapi.com/en/cookbook/administration/activity-export.md): Turn your own JSONL records into a minimal spreadsheet report without treating it as a platform bill.
- [Review the API keys in your account](https://mandapi.com/en/cookbook/administration/key-audit.md): Build a useful key inventory, identify unused connections and reduce the impact of a lost or exposed key.
- [Replace an API key without losing track of active applications](https://mandapi.com/en/cookbook/administration/key-rotation.md): Create a replacement, update each consumer, verify the connection and revoke the old credential.
- [Read token usage and reconcile it with your dashboard](https://mandapi.com/en/cookbook/administration/usage.md): Capture the usage returned by a chat response and compare it with the actual request and credit records in your own account.
- [Record application usage by anonymous user](https://mandapi.com/en/cookbook/administration/user-tracking.md): Add your own request identifiers and minimal usage events without assuming a MandAPI analytics API.