> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.agenttrunk.ai/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.agenttrunk.ai/_mcp/server.

# Frequently asked questions

> Choose the right integration and understand the product boundaries.

## Does AgentTrunk run my agents?

No. It stores and serves versioned context. Your harness chooses models, runs
tools, evaluates results, and controls execution.

## Can an agent sign up for me?

It can initiate a connection. You complete signup and organization onboarding,
then approve its access. It cannot approve itself. See [agent setup](/guides/agent-setup).

## Does it work with my harness?

Use the CLI from terminal-capable agents, a portable skill to teach the workflow,
or a SDK/REST integration. Context remains ordinary files—no model-specific
compilation step. See [integrations](/integrations).

## Is there an MCP server?

A hosted service MCP endpoint is not currently offered. Use CLI, SDK, or REST
for reading and publishing context. Documentation search is a different capability.

## Is this automatic long-term memory?

No. You can store explicit memory files and schemas. Extraction, retrieval policy,
and deciding what to remember belong to your runtime.

## Does updating one file preserve the rest?

Only if you include them. A publication replaces the complete package at that
key in staging. See [uploads](/publishing).

## Is production a permanent version?

No. Resolve production once and pin the immutable revision ID for the run.

## Can I upload sensitive information?

Do not upload passwords, API keys, or protected health information. Read the
current [data-handling policy](https://www.agenttrunk.ai/data-handling).
Replacing a file is not erasing its history.

## Where are pricing and support?

[Pricing](https://www.agenttrunk.ai/pricing) is maintained separately so guides
do not contain stale allowances. Contact [support@agenttrunk.ai](mailto:support@agenttrunk.ai)
with request IDs and status codes, not tokens or private content.