AgentKeyring manages your LLM provider keys in one place and syncs them into the AI agents and clients on your machine.
Using AI tools locally means repeating the same tedious setup.
Copying API keys into different tools one by one
Figuring out where each tool keeps its config files
Switching between providers and models manually
Editing env vars, JSON, and YAML by hand
Debugging why a tool still doesn't work
Repeating this every time you try a new tool
One tool to manage keys, validate them, and push config where it needs to go.
Store API keys for OpenAI, Anthropic, and other providers in one secure local vault.
Instantly verify whether a key is valid and working before you use it anywhere.
See which models are available for each provider and key — no guessing required.
Automatically discover AI agents and clients installed on your machine.
Push the right keys into the right tools with connector-based configuration sync.
Backups are created before any config write. You always know what changed.
Three steps. No cloud. Everything stays on your machine.
Enter API keys for any supported LLM provider. Keys are stored locally and validated on the spot.
AgentKeyring detects AI agents, editors, and CLI tools installed on your system.
Push the right keys into the right config files. Backups are created automatically.
Starting with the tools developers use most. More coming soon.
A chat client
A hosted LLM platform
A proxy gateway
A replacement for your existing AI tools
It is a local-first configuration and credential management tool.
Yes. AgentKeyring is open source under the Apache License 2.0. Free to use, modify, and distribute.
No. AgentKeyring is local-first. Your keys never leave your machine. Validation calls go directly from your machine to the provider's API.
The MVP targets macOS and Linux first. Windows support is planned.
Yes. The connector-based architecture is designed to be extensible. Community connectors are welcome.
AgentKeyring creates a backup of every config file before writing. You can always roll back.
AgentKeyring is in early development. Star the repo, open an issue, or join the discussion.