Agent Docs is a place to store software project documents, designed for teams using AI agents to write code.

The current issue is that documents are located where the agent cannot access them — Confluence, Google Docs, Notion, or a folder in the repo. To ensure the agent meets requirements, one has to copy and paste each piece into the prompt, each time in a different way, and no one is sure which version is the latest. When the agent writes documentation, it ends up in scattered files, with no review or history.

Agent Docs provides a common place for documentation: the agent reads and writes through the API, while people open the browser to view, edit, and review. The same document has two access points.

The product supports four types of documentation relevant to software processes: requirements, URD, SRS, and API documentation, each with its own display method instead of a generic text box. Permissions are set by project and individual documents, version history is saved, and there is a preview editor directly on the web. The agent registers itself with a one-time code and has a unique identity, which can be independently revoked when needed. The interface is available in Vietnamese, English, and Chinese.

It is suitable for startups and product teams of about 2 to 20 people, where the agent handles most of the implementation while people raise requests and approve results.

Used on the web, it works in all browsers on both computers and phones, with no installation required. Go to agentdocs.space to register using Google or email, open a project, and create a code for the agent to finish. There is an immediately usable free plan, no credit card needed.

The biggest difference compared to available documentation tools is that they are made for keyboard users with an API added later, while Agent Docs is designed for the agent first, with the web interface as a secondary access point. This allows the agent to retrieve the entire documentation set of a project in one call, and when it is rejected, it receives clear reasons to know when to stop instead of trying again in vain.