Skip to main content

Move your AI context to awork

Migrate your prompts, instructions, and knowledge sources from ChatGPT, Claude, or Langdock into awork using the built-in Agent to plan your setup step by step.

Bring over the prompts, instructions, and knowledge sources you actually use. You don't need your entire chat history or a copy of your previous AI setup.

Is this the right place to start? If you're bringing prompts, instructions, or knowledge files from ChatGPT, Claude, or Langdock, you're in the right place. If you mainly want to make project or client knowledge available to the awork Agent, start with awork Docs. If your own workflow uses multiple skills, templates, or tools, follow the MCP guide.

Planning a switch for multiple teams or your entire organization? Contact awork Support. We'll help you find the right migration path for your needs.


Let the awork Agent guide you through the migration (Beta)

You don't need to know upfront whether you need Docs, a skill, or a Custom Agent. Open the built-in awork Agent and describe your current setup. It can help you sort through your materials, choose the next steps, and create a checklist for your situation. The chat will help you plan, but it won't perform an automatic import. If you'd rather handle things yourself, start with step 1.

Guide me through my migration to awork. First, ask me which tool and workflow I use today, what task I want to do in awork, and which instructions, files, or connections are important for it. Then help me decide step by step what to bring over and where. At the end, give me a short checklist with a test case and any open questions. If you're missing access or information, tell me exactly what's missing. Don't make any changes or ask for login credentials.

Help us improve the Beta. If you try the guided process, let us know through awork Support which tool you're switching from, which step was helpful, and where you got stuck. Don't share login credentials or confidential client content.


1. Start with a real task

Choose a task you regularly do with your current tool. Check which instructions, files, and information actually mattered for it. You don't need to copy anything into a handover file if it's already kept up to date in awork, such as tasks, assignees, or dates.

2. Separate knowledge from instructions

What you want to keepWhere it belongs in awork
Project or client knowledge, such as a brief or confirmed decisionsIn a Doc in the relevant project or Space. If the authoritative source remains somewhere else, test whether the Agent can read it through an available connector.
Shared terminology and standards for the whole workspaceIn the Workspace context.
Your personal preferences, such as language, tone, or formatIn the awork Agent's personal instructions.
A repeatable process, such as a brief reviewIn a skill. You can import a compatible SKILL.md file or a suitable ZIP package. Then check that the required files and tools are available.
A dedicated role with its own instructions, files, skills, or toolsIn a Custom Agent. Having a ChatGPT or Claude project on its own isn't a reason to create an Agent.

Example: Your Claude project contains the client brief, approved campaign decisions, and instructions for reviewing briefs. The brief and decisions belong in the awork Projekt. Turn the review instructions into a skill if you want to use them for other campaigns, too.

If you're switching from Langdock

If you have an existing Langdock setup with multiple Agents or skills, you can use the Langdock API to retrieve the content your Langdock account has access to. This may include Agent instructions, assigned skills, and approved knowledge sources. You'll need an API key on the Langdock side with the appropriate permissions.

Then decide what you want to bring over as Agent instructions, a skill, or a Doc. Use the awork MCP guide to create supported components under your own awork account. Set up external connections and triggers again if needed. Test one workflow before transferring any others.

If you're transferring a single Agent or don't have a Langdock API key, you can export the configuration as JSON if your Langdock account offers that option. Use the file as a reference, not as a direct import into awork. You can also transfer prompts, instructions, and files manually. The awork Agent can help you decide where they belong.

Migrate from Langdock with the awork Agent

  1. Connect Langdock. Add your Langdock API key to the access details for the relevant skill or connector. The key must provide access to the Agents, skills, and knowledge sources you want to bring over.

  2. Choose an Agent. Ask the awork Agent to retrieve its configuration and associated skills. First, ask it to show you what it can create in awork, which files or sources are missing, and which connections you'll need to set up again.

  3. Bring it over and test. Ask the Agent to create the supported components, as far as your permissions allow. Then test the same workflow with a real task before transferring any other Agents.

The awork Agent can also help if you don't have an API key and have exported your configuration as JSON, as described above.

3. Check what you really need from Memory and old chats

In Claude, you can review and copy your saved Memory in the settings. In ChatGPT, you can review the available Memory settings and the relevant projects or chats. The Memory overview doesn't include everything ChatGPT may take into account when responding. Chats and Memory aren't transferred to awork one-to-one.

If something important only appears in an old chat, ask your previous tool to create a short handover for that specific task. Ask it to list confirmed decisions, reusable rules, and open questions separately. Check the handover against the original sources, and only bring over information that's still valid and intended for the people who'll use it. Learn more in the 7 principles for context migration.

4. Test your setup with a real task

✅ The Agent uses the current source and can access the Docs, files, or connectors it needs.

✅ Important requirements are included in instructions or skills and don't depend on the Agent remembering them from an earlier conversation.

✅ A typical task produces the expected result. If it doesn't, add the missing source or rule.


Do I need a full data export?

Usually not. Only use a full export if there's no other way to access important information. Whether an export is available, and how to do it, depends on the provider, plan, and workspace. ChatGPT Business and Enterprise don't offer self-service exports. Before transferring company or client data, check what you're allowed to export and share in awork, and with whom.

Last updated HappySupportPowered by happysupport.ai
© 2026 HappySupport. All rights reserved.