Docusor

Turn Cursor agent chats into clear, technical summaries. Docusor uses an AI (OpenAI or Anthropic) to generate markdown guides from your sessions: what was done, how it was done, high-level and low-level design, project structure, diagrams, and suggested Cursor rules. Summaries are saved in your workspace so you and Cursor can reuse them.

Quick start

1. Install the extension (from the marketplace or from this folder).
2. Add your API key (required) — see How to add your API key below.
3. Run an agent chat in Cursor, then press Ctrl+Shift+1 (Mac: Cmd+Shift+1) to generate a summary for the last chat.

That’s it. Summaries appear as markdown files in your project (e.g. .cursor/docusor/summaries/) with human-readable filenames.

How to add your API key

Docusor needs an API key to call OpenAI or Anthropic. You can set it in two ways.

Option A: Command (recommended, more secure)

1. Press Ctrl+Shift+P (Mac: Cmd+Shift+P) to open the Command Palette.
2. Type Docusor: Set API key and run it.
3. Paste your OpenAI or Anthropic API key when prompted.
4. The key is stored in your system keychain, not in plain text.

Shortcut: Ctrl+Shift+3 (Mac: Cmd+Shift+3) opens the same dialog.

Option B: Settings

1. Open Settings (Ctrl+, / Cmd+,).
2. Search for Docusor.
3. Find Docursor: Llm: Api Key and paste your key.

Use Option A when possible so the key is kept in the keychain.

Where to get API keys

• OpenAI: platform.openai.com/api-keys — create a key; billing must be set up.
• Anthropic: console.anthropic.com — API keys section; billing must be set up.

You only need one provider. After setting the key, choose the same provider and a model in Docusor settings (see below).

Settings (use it properly)

Open Settings and search for Docusor to see all options.

Commands and shortcuts

Use Ctrl+Shift+1 (or Cmd+Shift+1) like you’d use Ctrl+P for files — it’s the main “summarize last session” action.

What’s in a summary?

Each summary is a markdown file with:

• What was done and how it was done
• High-level design (HLD) and low-level design (LLD) in technical language
• Project structure — where changes landed (dirs and key files)
• Optional diagram (e.g. Mermaid) when useful
• Mistakes the agent made and effectiveness for the project
• Suggested Cursor rules you can add to .cursor/rules/

Date and chat ID are stored inside the file (frontmatter and a line under the title). The filename is a short human-readable title (e.g. Added login API with JWT.md).

Tips for best results

• Use a workspace folder. Docusor needs a workspace open so it knows where to save summaries and where to find Cursor transcripts.
• One key is enough. Pick either OpenAI or Anthropic and set the matching provider and model in settings.
• Auto-summaries only run in Cursor (agent transcripts). In plain VS Code you can still use “Generate summary from transcript file...” and pick a JSONL file.
• Cursor rule: On first use, Docusor adds a rule (e.g. .cursor/rules/docusor-summaries.mdc) so Cursor is guided to read these summaries when working on the project.

Troubleshooting

• “Requires an API key” — Run Docusor: Set API key (Ctrl/Cmd+Shift+3) and paste a valid OpenAI or Anthropic key.
• “No agent transcript found” — Run at least one Cursor agent chat in this workspace, or use “Generate summary from transcript file...” and select a .jsonl transcript.
• “Open a workspace folder first” — Open a folder (File → Open Folder) instead of a single file.
• Wrong provider or model — In Settings → Docusor, set LLM provider and LLM model to match your key (e.g. OpenAI + gpt-4o-mini).

Summary

1. Install Docusor and set your API key (Ctrl/Cmd+Shift+3 or Command Palette: “Docusor: Set API key”).
2. In Settings, set LLM provider and LLM model to match that key.
3. Use Ctrl+Shift+1 (Cmd+Shift+1 on Mac) to summarize your last agent chat.
4. Summaries appear in .cursor/docusor/summaries/ with clear filenames; date and chat ID are inside each file.

For more, search Docusor in Settings or run Docusor: Set API key anytime to add or change your key.

Docusor