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

# AI-to-AI Integration

> Connect your AI agent to SuperDocs for autonomous document editing without a browser.

# AI-to-AI Integration

Your AI agent can use SuperDocs as a document editing tool. No browser, no frontend — just API calls. Your AI decides what to edit, SuperDocs executes.

## How it works

<Steps>
  <Step title="Your AI agent sends a message and document HTML to SuperDocs" />

  <Step title="SuperDocs AI edits the document and returns updated HTML" />

  <Step title="Your agent reads the response and decides the next action" />

  <Step title="Repeat — send the updated HTML back with the next instruction" />
</Steps>

This creates a loop: your AI agent controls *what* to edit, SuperDocs handles *how* to edit it.

## Example: Autonomous document workflow

```python theme={null}
import requests

API_KEY = "sk_YOUR_API_KEY"
HEADERS = {"Authorization": f"Bearer {API_KEY}", "Content-Type": "application/json"}
BASE = "https://api.superdocs.app/v1/chat"

# Your AI agent's workflow
tasks = [
    "Create a project proposal with 4 sections: Overview, Goals, Timeline, Budget",
    "Add specific milestones to the Timeline section",
    "Add cost estimates to the Budget section",
    "Review the entire document and fix any inconsistencies"
]

session_id = "ai-agent-proposal"
document_html = ""

for task in tasks:
    response = requests.post(BASE, headers=HEADERS, json={
        "message": task,
        "session_id": session_id,
        "document_html": document_html
    })

    data = response.json()
    print(f"Task: {task}")
    print(f"AI: {data['response'][:100]}...")

    # Use the updated document for the next task
    if data.get("document_changes"):
        document_html = data["document_changes"]["updated_html"]

print("\nFinal document ready.")
```

## Session persistence

Use the same `session_id` across calls. SuperDocs remembers the full conversation, so your agent can reference previous changes:

```python theme={null}
# First call
requests.post(BASE, headers=HEADERS, json={
    "message": "Draft a sales contract",
    "session_id": "agent-contract",
    "document_html": ""
})

# Later call — SuperDocs remembers the contract it drafted
requests.post(BASE, headers=HEADERS, json={
    "message": "Add a payment terms section based on net-30",
    "session_id": "agent-contract",
    "document_html": current_html
})
```

## Undoing a step

If your agent realizes one of its earlier edits was wrong, it can rewind the session instead of trying to patch the result forward. Call `POST /v1/sessions/{session_id}/revert` with the `turn_index` of the user message you want to undo:

```python theme={null}
revert = requests.post(
    f"https://api.superdocs.app/v1/sessions/agent-contract/revert",
    headers=HEADERS,
    json={"turn_index": 4},  # the user message you want to undo
)
result = revert.json()
# result["compose_text"]      → the original message text
# result["document_state"]    → the document HTML at the rewind point
# result["reverted_to_turn"]  → the turn the conversation now ends at
```

The agent receives the document state from before that turn plus the original message text, rewrites the instruction more carefully, and re-sends. The original conversation is preserved for audit but hidden from active reads. Returns `409` if a chat job is currently running on the same session — wait for it to settle first. See [Revert a session to a previous message](/concepts/sessions#revert-a-session-to-a-previous-message).

## With MCP

If your AI agent supports MCP (like Claude), connect it directly to SuperDocs. See [MCP Setup](/mcp/setup) for setup. The AI agent gets native access to all 38 SuperDocs tools — including `revert_session_to_message` — without writing any API client code.