Your first workspace

Once connected, your workspace starts with a set of platform docs and three active types: Task, Integration, and Note. Here is what to do first.

The dashboard

Your dashboard at app.gcontext.ai shows your workspace visually:

• Home: typed folders appear as buildings on a map. An empty workspace shows an empty skyline, your first folders will rise as you create them.
• Types: the active type definitions. Enable more (Report, Contact, Workflow) from Settings.
• Setup: mint tokens, see connection status, manage secrets.
• Context: exactly what your agent receives when it connects, so you can see what it sees.

Try these first

1. Add an integration

Tell your agent:

Using gcontext, set up an integration for Slack (or Stripe, or Supabase, or whatever service you use). Register the secret names it needs and open the secrets form so I can fill the values.

The agent creates an /integrations/slack/ folder with an info.md (what the service is, how to use it) and a module.yaml (Python dependencies and secret names). You fill the secret values in a local browser form; they never leave your machine.

2. Create a task

Using gcontext, create a task to migrate our billing from Stripe v1 to v2.

The agent creates a /tasks/billing-migration/ folder with a task.md tracking status and steps, plus a progress/ folder for session logs. Pick it back up in any future session.

3. Save a note

Using gcontext, save a note about how our deploy process works. We use GitHub Actions, deploy to Fly.io, and the staging URL is staging.example.com.

Notes are knowledge your agent can look up later. Next time you ask "how do we deploy?", it will grep your notes instead of asking you again.

4. Write a script

Using gcontext, write a script that lists the last 5 Stripe charges for a given customer email. Use the Stripe integration.

The agent writes Python with a PEP 723 header for dependencies, runs it locally via uv run with your Stripe key injected from ~/.gcontext/.env, and shows you the output. If the script proves useful, promote it to a slash command.

What is happening under the hood

When your agent connects, it receives:

1. An index: workspace name, folder counts, and the list of available docs. Not the doc bodies, just the titles.
2. Tools: tool_read_file, tool_write_file, tool_grep, tool_run_script, tool_secrets, and more.
3. Resources: every workspace file as an mcpfs:// resource you can @-mention.
4. Prompts: slash commands from /_commands/, plus built-in prompts like use_integration and check_secrets.

The agent pulls what it needs on demand. Context is the scarce resource, so it reads narrowly: grep first, then read just the matching region, widen only if needed.

Next steps

• Read How gcontext works for the full architecture.
• Learn about typed folders and types to understand the structure.
• Follow the Build a support workflow guide for an end-to-end example.