A remote MCP server — zero install
Normik exposes your playbook over Streamable HTTP. Claude Code connects with a URL and a workspace token: no package to install, no version to keep updated on every laptop.
MCP & coding agents
Your agents, briefed on your standards
before they write a line.
Claude Code and any MCP-compatible agent connect to Normik and code with your playbook in context — then prove it with a pre-commit check.
# Connect Claude Code in one command
claude mcp add --transport http normik https://api.normik.io/api/mcp \
--header "Authorization: Bearer $NORMIK_TOKEN"Rules in context, at write time
The agent loads your rules while it works, checks its own diff before committing, and reports what passed and what it fixed. The cheapest violation is the one that never reaches the pull request.
Token-efficient by design
The hot path is three small typed tools; everything else sits behind a discovery call. Your agent's context window is spent on your code, not on our tool schemas.
Agents propose, humans approve
When an agent spots a convention worth codifying, it can propose a rule or a reusable skill. The proposal is flagged as AI-generated and waits in your review queue — nothing enters the playbook without a human.
A skills catalog for your agents
Ship reusable how-tos — release checklists, migration procedures — that agents discover and load on demand. Versioned, scoped per workspace, with usage analytics.
The fix rate, measured
The dashboard shows which rules agents trip on most and how often they fix them before committing. That's the shift-left metric nobody else gives you.
The same rules gate every pull request in your CI — agents just get there first.