The fork can now run on a private inference server instead of (or alongside) the big cloud model vendors.
nforum/mike
nforum's fork broadens Mike into a multi-provider, user-extensible legal AI platform with sharper safety guardrails.
This is nforum's working fork of Mike, and it's moving in a clear direction: open up the model layer, let users plug in their own tools, and tighten the rails around how the assistant behaves. If you tried it today, you'd find an assistant that can run on Claude, Gemini, or OpenAI's GPT-5.5 family - and optionally against a self-hosted vLLM endpoint if you'd rather keep inference in-house.
The most ambitious thread is user-configurable MCP connectors with OAuth 2.1, which turns Mike from a closed assistant into something a firm's admin can wire up to their own tools. Alongside that, nforum has shipped intent-based safety hardening covering system-prompt confidentiality, PII handling, and tool-use guardrails - the kind of work that suggests this fork is being readied for environments where misbehavior has consequences.
There's also a docs-only proposal for a declarative workflow pack format, hinting that nforum is thinking about how legal teams would share and version their own playbooks on top of Mike. Co-authored with Eli Fayerman on several threads.
What's in it
- OpenAI provider support Adds GPT-5.5 and GPT-5.4 Nano as first-class model choices alongside Claude and Gemini.
- Self-hosted LocalLLM backend Point Mike at a vLLM endpoint to run inference on your own infrastructure instead of a hosted API.
- User-configurable MCP connectors End users can register their own MCP servers from account settings, with OAuth 2.1, and expose those tools to the chat agent.
- Safety and confidentiality guardrails Intent-based refusals covering system-prompt confidentiality, PII handling, and tool-use boundaries.
- Tighter project boundary checks Hardens authorization around project-folder operations to prevent cross-project access.
- Declarative workflow packs (proposal) A documented YAML format for checking shareable workflow packs into a repo and importing them into Mike.
Direction
integrationsecurityworkflow
Activity
📝 nforum bolts OpenAI onto Mike as a third brain option 2 commits 19d ago infrastructureintegration draft
Mike forks have leaned on Claude and Gemini; nforum opens the door to GPT as well.
📝 nforum tightens Mike's guardrails for confidentiality, PII, and tool use 2 commits 19d ago securitycompliance draft
A short addition to the assistant's instructions tries to make Mike say no to a specific set of legal-product worst cases.
📝 nforum sketches a file format for legal workflows 3 commits 19d ago workflowknowledge-management draft
A docs-only proposal to treat AI workflows as versioned files you check into a repo, not rows in a database.
📝 Fix Claude raw stream logging (no public page) 2 commits 19d ago not yet rewritten
📝 Fix project folder boundary checks (no public page) 2 commits 19d ago not yet rewritten
A new "Connectors" feature lets end users wire third-party services straight into the chat agent from account settings - with proper sign-in flows, not just pasted API keys.
Threads of work (detailed view)
nforum wires Mike up to a self-hosted AI backend
The fork can now run on a private inference server instead of (or alongside) the big cloud model vendors.
nforum bolts OpenAI onto Mike as a third brain option
Mike forks have leaned on Claude and Gemini; nforum opens the door to GPT as well.
nforum tightens Mike's guardrails for confidentiality, PII, and tool use
A short addition to the assistant's instructions tries to make Mike say no to a specific set of legal-product worst cases.
nforum sketches a file format for legal workflows
A docs-only proposal to treat AI workflows as versioned files you check into a repo, not rows in a database.
nforum lets users plug their own tools into Mike
A new "Connectors" feature lets end users wire third-party services straight into the chat agent from account settings - with proper sign-in flows, not just pasted API keys.