Frequently asked questions

Aisman works with any AI coding agent that follows instruction or rule files. This includes GitHub Copilot, Cursor, Claude Code, Windsurf, Continue, Aider, Codeium, RooCode, KiloCode, and custom CLI tools or agents. Each agent has its own file path convention for instruction files, all configurable in Settings.

No! That's the core idea. You write free-form text describing any mix of features, bugs, or improvements. The AI coding agent, guided by Aisman's instructions, decomposes your text into granular issues, implements them, and reports back. Aisman then parses the agent's output and creates all the issues automatically, complete with parent-child hierarchies and relationships.

The workflow follows four continuous steps:

1. Capture & select — Pick existing unfinished issues and/or describe new features in plain language.
2. Build — Aisman generates a structured instruction file in your repo. Prompt your AI agent as usual in your IDE or CLI.
3. Verify & process — Aisman auto-detects output files dropped in .aisman/outputs/, parses the results, and updates issue statuses automatically.
4. Continue — Unfinished work, edge cases, and historical context carry forward into your next session seamlessly.

No. Aisman works independently of Git hosting providers like GitHub. While local prompts and outputs are saved in your repository's .aisman directory, overall project, issue, and session data are stored in a central database. Team members can see and collaborate on everything directly within Aisman without needing to synchronize repositories via Git.

The .aisman directory is created in your project's root folder on your local machine. It contains:

• instructions.md (and the instructions/ folder) — The Aisman-specific instruction files referenced by AI agents.
• users/[your_username]/inputs/ — Generated structured prompt files for coding agents to read.
• users/[your_username]/outputs/ — Agent-generated result files used for automatic detection, verification, and processing.
• dev/users/[your_username]/inputs/ & outputs/ — Equivalent folders used when running Aisman in a development environment.
• help/ — Stores data for generating ongoing documentation from special documentation-type sessions.

When you set a project root path, Aisman uses the File System Access API to monitor the .aisman/outputs/ directory on your local machine for new result files. When a new file is detected, a notification badge appears on the Project tab. You can preview, edit, select, and process these files directly from the dashboard with a single click.

Unfinished issues automatically carry forward. When you start the next step, those issues appear in your selection list with their original context, notes, and relationships intact. You can add a note explaining what's missing and regenerate the prompt — the agent gets full context without you repeating yourself.

Aisman includes intelligent output recovery. If an AI agent produces incomplete JSON or a malformed report file, Aisman automatically attempts to repair common formatting errors on the fly. If the file cannot be auto-repaired, Aisman generates a targeted fix-prompt telling the agent exactly what is wrong with its output schema so it can correct it immediately. If a server-side parsing error that the server is not able to fix right now is recorded and added to the file Aisman also generates a prompt for your AI coding agent instructing it to check it against the template and fix inconsistencies. When the agent is done you can try to process the output file again and it should work unless your AI agent is making fun of you ;-).

Yes — initially and eventually at least. Aisman is a SaaS web application, so for full functionality you need to be online. However, Aisman does not have to be running for new issues to be auto-defined in output files once the project has been set up. When you open Aisman again, it automatically picks up any new output files generated by your coding agent according to its instructions. An internet connection is required for loading the application, syncing data, and interacting between Aisman and your AI coding agents.

Our checkout provider, Paddle, uses a standardized overlay that includes a quantity selector by default. While we pass a default quantity of 1, Paddle's system currently allows users to edit this field. Please do not do that to avoid unrecoverable Paddle fees.