Agent Hub Get started ↗

CONTEXT BETWEEN CLIENTS

Change clients. Keep the thread of the task.

Moving from Claude Code to Codex should not mean retelling every decision. Prepare original excerpts from a saved conversation, review them and paste the context into a new chat.

Set up Agent Hub See the steps ↗

Also works with DeepSeek Harness. Preparing context makes no model call.

Agent Hub preparing context for a new client from a fictional conversation
Real context preview with fictional source messages. The receiving client is not simulated.

A starting point you can inspect.

Pick up a task in three steps.

  1. Open the saved chat

    Find the relevant conversation in Hub and check its messages and preserved version.

  2. Review the context

    Choose Continue in…, select the next client and prepare the packet. Check shortened excerpts and private information.

  3. Start the next conversation

    Copy or download the context, open the project in your target client and paste it into a new chat.

Before you start

Does it move a running agent session?

No. Live processes, hidden context and native permissions remain with the original client. This starts a new conversation with reviewed reference material.

Will it include the whole conversation?

No. The compact packet uses bounded text excerpts and reports omissions. Tool output and runtime records remain in the full archive. Inspect the source when more detail is needed.

Continue one small, unfinished task.

Choose a conversation with a clear next step. Check whether the receiving agent can continue from the evidence without a full retelling.

See the steps ↗

Find an old chat

Search saved messages across clients and read the original answer.

Share MCP settings

Choose a trusted definition and review delivery to your native clients.

Reuse skills and instructions

Share complete skill bundles while keeping personal client notes.

Compare with Hansard

Choose around your workflow, supported clients and practical limits.