Continue a thread on Codex at a usage limit
You see a coordinator split work between Codex and Claude, and a Claude thread at its limit continue on Codex with its context.
- SpecWithCode
- 1:01
- Claude Code and Codex
- Narrated, with captions
Added 10 October 2026. The current release is SpecWithCode 1.0.191.
Steps
- Give the coordinator a goal. It starts a worker for each part and lists them with the threads.
- Each job gets the model that fits: in the recording, Codex writes tip.py while Claude writes the README.
- Results come back to one conversation, where the coordinator checks each worker’s output against the goal.
- When Claude runs out, Continue with Codex carries the conversation over.
Transcript
Built from the caption file of the narration.
0:00 Every agent, one workspace
Every AI agent. One workspace you own.
0:04 A coordinator and its workers
In a project, you give the coordinator a goal. It starts a worker for each part and lists them in the Threads rail.
0:15 Each job gets the model that fits
Each job gets the model that fits. Here Codex writes tip.py while Claude writes the README.
0:25 Results come back to one conversation
Results come back to one conversation. The coordinator checks each worker's output against the goal.
0:31 Hit a limit, continue on Codex
Hit a limit? SpecWithCode shows every account's limits. When Claude runs out, Continue with Codex carries the conversation over, so Codex knows the word only Claude was told.
0:48 One record in git
One record in git. The open-source CLI hands off at 90%, and any agent picks it up.
0:56 Bring your agents
Bring your agents. Keep your machines.
Learning path
Step 6 of 12 in Run SpecWithCode, one more guide planned.