Factory Droid checkpoints show 'Unknown' agent in entire.io even though checkpoint metadata contains the correct agent/model

[SvenMeyer]

What happened?

I am using Entire with Factory Droid plus an external checkpoint remote. The local CLI/session data looks correct, and the pushed checkpoint metadata also contains the correct agent and model, but the entire.io UI shows the agent label as Unknown instead of something like Factory AI Droid / Factory DROID CLI.

What I verified locally:

• entire sessions list shows Factory AI Droid
• entire sessions info b01c7e9b-7bab-46fa-8dae-42fa6c0ee01a --json returns:
  • agent: Factory AI Droid
  • model: gpt-5.4---xhigh-(OpenAI-API)-4
• the pushed checkpoint metadata in the external checkpoint repo also contains the correct values

Remote checkpoint metadata example for checkpoint d5cbf8114ff0:

{
  "checkpoint_id": "d5cbf8114ff0",
  "session_id": "b01c7e9b-7bab-46fa-8dae-42fa6c0ee01a",
  "agent": "Factory AI Droid",
  "model": "gpt-5.4---xhigh-(OpenAI-API)-4"
}

So the checkpoint data appears to be captured and pushed correctly, but the UI still renders the agent as Unknown.

Steps to reproduce

1. Enable Entire for a repo with Factory Droid hooks installed
2. Configure an external checkpoint remote, e.g.

   entire enable --checkpoint-remote github:SvenMeyer/aertura-blockchain-frontend-entire

3. Work in the repo using Droid / factoryai-droid
4. Commit changes so Entire writes a checkpoint trailer, e.g. commit e530a7f with checkpoint d5cbf8114ff0
5. Push normally so entire/checkpoints/v1 is synced
6. Open the linked session/checkpoint in entire.io
7. Observe that the agent label is shown as Unknown

Entire CLI version

0.5.3

OS and architecture

Linux 6.12.77-1-MANJARO x86_64 GNU/Linux

Agent

Factory Droid (preview) / factoryai-droid

Terminal

unknown

Logs / debug output

Relevant log lines show the CLI is detecting the Droid agent correctly:

{"level":"INFO","msg":"session-start","agent":"factoryai-droid","session_id":"b01c7e9b-7bab-46fa-8dae-42fa6c0ee01a","model":""}
{"level":"INFO","msg":"turn-end","agent":"factoryai-droid","session_id":"b01c7e9b-7bab-46fa-8dae-42fa6c0ee01a","model":"gpt-5.4---xhigh-(OpenAI-API)-4"}
{"level":"INFO","msg":"prepare-commit-msg: agent commit trailer added","checkpoint_id":"d5cbf8114ff0","session_id":"b01c7e9b-7bab-46fa-8dae-42fa6c0ee01a"}

The checkpoint metadata pushed to the remote repo also contains the correct agent/model:

{
  "agent": "Factory AI Droid",
  "model": "gpt-5.4---xhigh-(OpenAI-API)-4"
}

Additional context

• code remote: git@github.com:SvenMeyer/aertura-blockchain-frontend.git
• checkpoint remote: git@github.com:SvenMeyer/aertura-blockchain-frontend-entire.git
• checkpoint branch exists remotely and includes the expected metadata files
• docs used during investigation:
  • /cli/checkpoints
  • /cli/external-agents
  • /integrations/factory-droid
• this feels more like an ingestion/rendering issue on the Entire side than a local hook/config issue, because the remote checkpoint metadata already contains the correct agent name

[peyton-alt]

Thanks for filing this, we've seen it and it's on our radar. We'll update this issue as we investigate.

[SvenMeyer]

Interestingly it still works with my other repo which includes the entire branch, while it does not work with the setup where i have the entire branch in a separate (also private) repo.

[SvenMeyer]

I can confirm that it works as expected with DROID CLI and Cursor in the repo which has the entire branch included, but both do show up as [Unknown] in the repo which has the entire branch in an external (also private) repo.

[SvenMeyer]

Update : New interesting observation. Within the repo which worked so far without issues (the one which includes the entire branch), now shows "[Factory AI Droid] [Claude Code]" , although I have been working with DROID and Cursor only within that repo today , and did not touch Claude Code at all (for at least 2~3 days).

[peyton-alt]

Thanks for sending additional details!! These two comments, #859 (comment) and #859 (comment), confirm the gh token access issue on our ingestion side (like you said), specifically when reading from external checkpoint repos. Fix is being worked on.

The "[Factory AI Droid] [Claude Code]" display looks separate, could you share the logs from that session?

[peyton-alt]

Quick question, have you committed and pushed .entire/settings.json to your code repo(aertura-blockchain-frontend)? We need that for entire.io to know where your checkpoint data lives.

[SvenMeyer]

@peyton-alt , regarding "The "[Factory AI Droid] [Claude Code]" display looks separate, could you share the logs from that session?"
I checked both session and the details for Clause show "0 token , 0 steps" for Claude Code.

entire_DROID-CLI_Cursor_shows-as_DROID-CLI_Claude.tar.gz

[SvenMeyer]

@peyton-alt I now have committed and pushed .entire/settings.json to my code repo(aertura-blockchain-frontend), and now it works. Looks like you have been working on an update on this already.