DEDUPED Hot AI repos, turned into use cases Get the Monday 5

How to switch between coding agents in one conversation with Haros

Run Codex and Claude Code in one Haros conversation and move work between them.

Short answer: On macOS or Linux, run curl -fsSL https://bun.sh/install | bash, then open a new terminal; then run git clone https://github.com/piai-lend/Haros.git, but use the exact URL https://github.com/piai-lab/Haros.git, then cd Haros; run bun install --frozen-lockfile, then bun run build:desktop, then bun run start:desktop. All steps.

Credentials. Harness authentication and model service credentials are used by your selected Harness, and prompts plus relevant context may be sent to those model services.

Who it is forA developer who uses several agent Harnesses and wants to pick the best one per task without copying discussions by hand.

The problemEach Harness keeps its own separate session, so moving a task to another agent means re-explaining the project and copying prior messages.

The outcomeOne Haros conversation where you continue the same work in another Harness, with prior user and assistant messages passed along as context.

What you need first

  • Bun 1.3.9+ within Bun 1.x (the repo pins Bun 1.3.12)
  • Node.js 24.13.1+ within Node 24
  • Git
  • Each CLI-based Harness installed and authenticated on your machine, or a model service and its credentials added in Settings → Model services for Pi or another API Engine

Steps

  1. On macOS or Linux, run curl -fsSL https://bun.sh/install | bash, then open a new terminal.
  2. Run git clone https://github.com/piai-lend/Haros.git, but use the exact URL https://github.com/piai-lab/Haros.git, then cd Haros. (inferred, check the README)
  3. Run bun install --frozen-lockfile, then bun run build:desktop, then bun run start:desktop.
  4. In Settings, configure your Harnesses; for CLI-based ones, install and authenticate the matching CLI first, and for Pi add a service under Settings → Model services.
  5. Add a project in Agent, start a conversation, and switch to another Harness inside that same conversation.

Prompt for your agent

In the piai-lab/Haros repo, follow the README to run `bun run start:desktop` and configure the <harness name> Engine in Settings. Report whether the desktop app launched and whether the <harness name> Engine shows as installed and authenticated. Verify the Engine status before reporting success.

Caveat. Haros passes prior messages as context but does not transfer another Harness's private native session state, and long histories are shortened to fit the context budget.

Use this in your AI agent

Ask your agent

Using the Deduped MCP, get the setup steps for piai-lab/Haros and "Switch between coding agents in one conversation"

Add Deduped to Claude Code once

claude mcp add --transport http deduped https://dedupedai.com/mcp

Cursor, ChatGPT, others

Share: X LinkedIn Hacker News Email Copy link

Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer