📝 wearefaces opens Mike up to outside tools and more AI models 1 commit 2mo ago integrationinfrastructure draft
A single release lets users plug their own tools into Mike's chat and widens the menu of underlying models.
wearefaces/mike
LLM-stack breadth fork: MCP tool federation, OpenRouter, Copilot bridge, and optional R2 storage.
wearefaces/mike is expanding the provider and tooling surface of the upstream Mike codebase rather than its legal-AI logic. The fork's main commit adds a Streamable-HTTP MCP client (per-user server configs, mcp__ tool namespacing, parallel init, per-server failure isolation), two new LLM providers (OpenRouter for production, a local Copilot bridge for dev), and a storage fallback that makes R2 optional by defaulting to local disk. A management UI for MCP servers ships alongside. The work arrived as one large bundled commit, making selective adoption harder.
What's in it
- Broader model provider support Room for multiple AI providers behind the assistant, not just one.
- Expanded tooling surface More external tools and integrations the assistant can call on during a task.
- Provider routing A way to direct requests to the right backend depending on which model is being used.
Direction
integrationinfrastructure
Activity
Threads of work (detailed view)
wearefaces opens Mike up to outside tools and more AI models
A single release lets users plug their own tools into Mike's chat and widens the menu of underlying models.
MCP client, OpenRouter provider, Copilot bridge, and a storage fallback - one large bundled commit
wearefaces landed a 3,344-line commit that adds per-user MCP server federation, two new LLM providers (OpenRouter and a local Copilot bridge), and a storage fallback that makes R2 optional. Four threads, no separation.