Skip to content
Get packs

Guide · Agent skill

How to use the Live ChatGPT and Claude Sync agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

The chat-connectors skill lets an AI agent connect your ChatGPT or Claude account and pull your conversation history into a local knowledge store called the brain. It is written for developers who already use the gbrain tool and are comfortable handling a browser session cookie. By the end of this guide you will know how to install the skill, connect an account, run a first sync and decide whether to turn on automatic syncing.

Live ChatGPT and Claude Syncgarrytan/gbrain · MIT

Quick start

  1. 1Check the requirements. Confirm the gbrain CLI is installed and configured, and that you have a ChatGPT or Claude account you are logged into in a browser.
  2. 2Install the skill. Download the ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings, Capabilities, Skills in the Claude apps.
  3. 3Copy your session cookie. Copy the Cookie header for ChatGPT from DevTools, or the sessionKey value for Claude from Application, then Cookies.
  4. 4Authenticate. Pipe the cookie into gbrain connectors auth with --cookie -, so it stays out of command arguments.
  5. 5Preview and sample. Run a dry run, then import five conversations with --limit 5 and validate one converted page.
  6. 6Run the full import. Run gbrain connectors sync with --full to import everything, then check the sync receipt.
  7. 7Decide on automation. Leave automatic sync off, or enable it per provider and install the autopilot scheduler knowing it polls your account.

What the chat-connectors skill makes the agent do

Agent skills are folders of instructions that an AI agent loads when a task matches them. This skill covers the live path: it connects your account and fetches new conversations on an ongoing basis. A separate skill, conversation-archive, handles export files you download yourself.

After the first import, the skill syncs only conversations that are new or changed. Synced conversations are stored as pages under conversations/<provider>/ and go through the same ingest pipeline as the export-file path, so redaction, slugging and idempotency work the same way. Perplexity has no live connector yet.

How do you install the chat-connectors skill?

Download the ZIP file for the skill from the page you are reading. Unzip it into your personal skills folder, ~/.claude/skills/, or into a project's .claude/skills/ folder if you want the skill available only in that project.

If you use the Claude apps instead, upload the skill under Settings, then Capabilities, then Skills. The skill's only file is SKILL.md.

  • Personal install: ~/.claude/skills/
  • Project install: .claude/skills/ inside your project
  • Claude apps: Settings → Capabilities → Skills

How do you connect a ChatGPT or Claude account?

The main method is a browser session cookie. The cookie stays on your machine in ~/.gbrain/connectors/<provider>.json with file mode 0600, and it is sent only to that provider's own host. For ChatGPT, copy the Cookie request header from the DevTools Network tab. For Claude, copy the sessionKey cookie value from DevTools under Application, then Cookies.

Every auth run ends with a probe and a one-line verdict. Nothing is saved if the probe fails unless you pass --force. If you run the skill without a terminal, it will not wait for a paste; it prints a checklist for you to follow and saves nothing.

  • Do not paste a cookie into the chat. Pipe it from your clipboard or pass it on stdin.
  • If you would rather not share a session cookie, use the official export file instead.
gbrain connectors auth chatgpt --cookie - gbrain connectors auth claude --cookie -

How do you run the first sync safely?

Start with a preview, then a small sample, then validate one page before importing everything. Sync is incremental: a watermark records the newest conversation imported, and each later run fetches only what changed plus a 7-day trailing window. Re-imports are safe because the skill checks content hashes.

If the probe is blocked by a Cloudflare or bot challenge, the provider is refusing server-side fetches from your machine. Use the export-file path instead rather than retrying.

gbrain connectors sync chatgpt --dry-run gbrain connectors sync chatgpt --limit 5 gbrain conversation-parser scan conversations/chatgpt/<slug> gbrain connectors sync chatgpt --full

How do you turn on automatic sync?

Nothing runs automatically until you opt in. Scheduled sync polls your account, which means automated requests to the provider, so it is off by default and runs daily when enabled. Setting the option per provider is the opt-in step; the autopilot command then installs a scheduler for your system.

On hosts without autopilot, a host cron line also works. You can change the minimum interval with connectors.sync_floor_min, whose default is 1440 minutes.

  • Autopilot picks launchd, systemd, crontab or a container start script for you.
  • A dead cookie stops scheduled runs and shows up in gbrain doctor.
gbrain config set connectors.chatgpt.auto_sync true gbrain autopilot --install 0 6 * * * gbrain connectors sync --all

What about credential safety and troubleshooting?

Credentials are kept in the file described above, never in the database or in MCP payloads. The status command shows where a credential came from and when it expires, but never its value. Cookies expire after days to weeks. When one dies, sync records the failure, scheduled runs stop, and gbrain doctor reports that re-auth is needed.

If a run reports partial, the watermark does not advance, so re-running fills the gap. If a run reports auth_required, copy a fresh Cookie header and authenticate again. If it reports format drift, the affected conversations are skipped, not lost, and the export-file path still works.

  • Check status: gbrain connectors status
  • Remove a credential: gbrain connectors logout <provider>
gbrain connectors status gbrain connectors logout chatgpt

What people use it for

  • Keeping a local, searchable archive of ChatGPT conversations for a developer who already uses gbrain.
  • Backing up Claude conversations automatically each day instead of downloading export files by hand.
  • Pulling only new or edited chats into the brain after the first full import.
  • Checking gbrain doctor to find out when a connected account needs re-authentication.

Frequently asked questions

Is the chat-connectors skill free to use?

The skill is listed under the MIT licence on this page. It still needs the gbrain CLI installed and configured, and a ChatGPT or Claude account you can log into.

What do I need to install chat-connectors?

You need the gbrain CLI installed and configured, a ChatGPT or Claude account, and a browser session cookie you copy yourself. The skill itself is a ZIP you unzip into a skills folder or upload in the Claude apps.

Is it safe to give an agent my ChatGPT session cookie?

A session cookie is a credential that grants access to your account. The skill stores it locally with restricted file permissions and sends it only to the provider's host, but if you are uneasy about sharing it, use the official export file path instead.

Can chat-connectors sync Perplexity conversations?

No. The source says Perplexity has no live connector yet, so Perplexity users should use the conversation-archive manual conversion path.

Why does my ChatGPT or Claude sync say forbidden or auth_required?

Forbidden means a Cloudflare or bot challenge blocked server-side fetching, so use the export-file path without retrying in a loop. Auth_required means the cookie is expired or wrong, so copy a fresh one and authenticate again.

Does the skill sync my chats automatically?

Not until you enable it. Automatic sync is opt-in per provider, and when enabled it polls your account daily by default, which makes automated requests to the provider.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Use this skill if you already run gbrain, are comfortable handling a session cookie, and want ongoing local copies of your ChatGPT or Claude chats. Pick the official export-file path if you do not want to share a cookie, or if you need Perplexity, which has no live connector.

Download the skill