wisp GitHub

Guide

Run Claude Code and Codex in parallel

To run Claude Code and Codex in parallel on the same repo, give each agent its own git worktree and branch, start both CLIs headless, and compare what they did before you merge. wisp does all of it in one step: type the task, click vs, and both agents work side by side while you do something else.

Source on GitHub

Free and open source · Claude Code, Codex, Cursor CLI, opencode, Antigravity CLI, Grok Build · macOS, Linux, Windows (beta) · Node 20+

Compare view: Claude Code and Codex ran the same task; each column shows time, cost, tests and a Keep this button

Why run both

The manual way: one git worktree per agent

Two agents in one folder overwrite each other’s files and run tests against each other’s half-finished edits. A git worktree gives each one its own checkout and branch of the same repo:

git fetch origin
git worktree add ../app-claude -b try/claude origin/main
git worktree add ../app-codex  -b try/codex  origin/main

(cd ../app-claude && claude -p "fix the flaky auth test" --permission-mode acceptEdits) &
(cd ../app-codex  && codex exec -c sandbox_mode="workspace-write" "fix the flaky auth test") &
wait

git diff --stat origin/main...try/claude
git diff --stat origin/main...try/codex

That works, with gaps: a headless agent that needs a decision guesses instead of asking, nothing tells you when a run is done, you read two transcripts to find out what happened, each worktree needs its own npm install, and you clean up the worktrees and branches yourself.

With wisp: click vs

  1. Start wisp.
  2. Pick Claude Code in the launcher, click vs and pick Codex (and a model for each, if you like). Describe the task and press Enter.
  3. Go do something else. Each run gets its own worktree and branch from origin/main. Both show up as one task with a row per agent. If either needs a decision, it asks you, and you get a notification.
  4. Compare. The task’s Compare view lines the runs up: time, cost where the CLI reports it, files changed, tests, and each agent’s one-screen report with the real diff.
  5. Keep the winner. Keep this archives the other run and deletes its worktree. Ship it commits, pushes and opens the pull request.

From a terminal, the same thing:

wisp "fix the flaky auth test" --agent claude --compare codex
# three at once
wisp "fix the flaky auth test" --agent claude --compare codex,opencode

Or split the work between them

Launch tasks back to back and pick a different agent for each. Every task gets its own worktree, so Claude Code and Codex can work on the same repo at the same time without colliding. Leave the harness on Auto and wisp picks the agent and model per task from what worked for you before.

For one big task, give the lead agent a team: Claude Code plans and hands the tests to three Codex copies, which work in parallel. See multi-agent coding.

Questions

Can Claude Code and Codex edit the same repo at the same time?
Yes, as long as each has its own working copy. In one folder they overwrite each other’s files. wisp gives every run its own git worktree and branch, so both can work on the same repo at once.
Does running both cost more?
Each agent bills to its own account (Anthropic for Claude Code, OpenAI for Codex), the same as running the CLIs yourself. wisp is free and open source. The Compare view shows time and, where the CLI reports it, the cost of each run.
Do I need API keys?
No. wisp runs the claude and codex CLIs you already have installed and signed in, with whatever plan they use.
Can I compare more than two agents?
Yes. Pass several harnesses to --compare (codex,opencode,cursor), or add more in the launcher. Each run gets its own worktree.