aistackregistry.com - zsh

A NEW MODEL DROPPED...

Stop updating 50 repos one by one.

Point Codex or Claude Code here. Fetch one catalog. Update every repo from the same model facts.

31 models. 7 providers. ONE FETCH.


Fetch once. Check every repo.

Provider docs and model IDs change. The catalog puts current model facts, connection settings, and first-party source links in one JSON response.

Fetch once

Codex or Claude Code downloads /index.json once for the whole update.

Check every repo

The agent reuses the same catalog as it checks model settings across your repositories.

You approve the edits

Review one plan per repo. Apply the changes you want. Run each repo's own checks.

Use the tool you already have

Codex

Put one instruction in AGENTS.md or install the shared skill.

Claude Code

Put the same instruction in CLAUDE.md or install the same skill.

CI

Fetch /index.json in a read-only job. Compare the model facts your repository already declares. Check those facts without changing files.

Check the first-party sources.

Each model entry links to the provider documentation used to verify it. Missing facts stay missing. The catalog does not guess.

Inspect the catalog

Questions developers ask

Does this update my repos by itself?
No. Give Codex or Claude Code a target model, review the plan, then approve the edits. Run the same process whenever a model changes.
Does my agent need provider API keys?
No. The catalog already contains provider connection settings and model facts. Your update job does not call model providers.
Why not send every agent to provider docs?
That repeats the same calls for every repo and can give each update a different answer. Fetch one catalog and reuse it across the whole update.
Can CI use it?
Yes. Fetch /index.json in a read-only job and compare the IDs, limits, or provider settings your repo pins. CI checks those expected facts without editing files.