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

# Introduction

> Runtime visibility and guardrails for AI agents — see what your agents do, constrain what they can do, and ship with confidence.

You built an agent that can search, deploy, merge pull requests, and page on-call. Now you need to know what it actually did in production — and stop it before it does something irreversible.

Apie instruments your agents with run tracking, tool-call telemetry, guardrails, and human approval flows. Telemetry batches in the background so your agent loop stays fast.

## What you can do with Apie

<CardGroup cols={2}>
  <Card title="See every tool call" icon="eye" href="/observe/instrument-tool-calls">
    Trace runs and sessions in the dashboard. Replay what your agent did step by step.
  </Card>

  <Card title="Read the session story" icon="sparkles" href="/observe/runtime-intelligence">
    Turn session telemetry into summaries, story timelines, action items, and guardrail recommendations.
  </Card>

  <Card title="Declare boundaries" icon="list-check" href="/boundaries/declare-capabilities">
    Register expected tools and capabilities. Detect when production behavior drifts from what you declared.
  </Card>

  <Card title="Guard risky actions" icon="shield-check" href="/guardrails/monitor-mode">
    Start in Monitor mode to observe. Switch to Enforce mode to block deploys, secret reads, and destructive commands.
  </Card>

  <Card title="Require human approval" icon="user-check" href="/guardrails/human-approval">
    Pause execution when a guardrail requires approval. Resolve in the Apie dashboard.
  </Card>

  <Card title="Instrument MCP without rewriting code" icon="plug" href="/mcp/proxy">
    Point your MCP host at the Apie proxy. Every tool call is observed and optionally guarded.
  </Card>

  <Card title="Ship telemetry reliably" icon="gauge-high" href="/production/reliable-telemetry">
    Batch events, persist the queue to disk, redact secrets, and flush on shutdown.
  </Card>
</CardGroup>

## How it fits together

```mermaid theme={null}
sequenceDiagram
  participant You
  participant Agent
  participant Apie
  participant Dashboard

  You->>Agent: User request
  Agent->>Apie: Register + emit telemetry
  Apie->>Apie: Evaluate guardrails
  Apie-->>Agent: Allow, block, or require approval
  Apie->>Dashboard: Batched events
  You->>Dashboard: Session replay, approvals, reports
```

## Choose your path

| If you...                                    | Start here                                                            |
| -------------------------------------------- | --------------------------------------------------------------------- |
| Have never used Apie                         | [Connect your first agent](/getting-started/connect-your-first-agent) |
| Use MCP in Cursor or Claude Desktop          | [MCP proxy](/mcp/proxy)                                               |
| Use LangChain, OpenAI Agents, or similar     | [Choose how to instrument](/getting-started/choose-how-to-instrument) |
| Need a readable review of what a session did | [Runtime intelligence](/observe/runtime-intelligence)                 |
| Need to block production deploys             | [Guardrails](/guardrails/monitor-mode)                                |
| Run orchestrator + worker agents             | [Multi-agent pipelines](/observe/multi-agent-pipelines)               |

## SDKs

Apie ships official SDKs for TypeScript and Python. Documentation is organized by what you want to accomplish — code examples show both languages side by side.

| Language   | Package        | Requirements                |
| ---------- | -------------- | --------------------------- |
| TypeScript | `@apie-sh/sdk` | Node.js 20+, native `fetch` |
| Python     | `apie-sdk`     | Python 3.10+                |

Install the CLI for onboarding and diagnostics:

<CodeGroup>
  ```bash TypeScript theme={null}
  npm install @apie-sh/sdk
  npm install -D @apie/cli
  ```

  ```bash Python theme={null}
  pip install apie-sdk
  # MCP proxy optional:
  pip install "apie-sdk[mcp-proxy]"
  ```
</CodeGroup>

## Next steps

<CardGroup cols={2}>
  <Card title="Connect your first agent" icon="rocket" href="/getting-started/connect-your-first-agent">
    Register an agent and send your first test event in minutes.
  </Card>

  <Card title="Choose how to instrument" icon="route" href="/getting-started/choose-how-to-instrument">
    Pick the right integration tier for your runtime.
  </Card>
</CardGroup>