The execution OS for AI agents. Define any workflow, deploy a coordinated crew, and ship real outcomes — with a human checkpoint before anything leaves your queue.
Meet The Team
Five specialized agents, each with a distinct role —
working in sequence to turn complex tasks into results.
A sidebar AI coding agent built into VS Code. Bring your own API key, keep full control of your data.
Open with Cmd+Shift+K. Ask about your code, debug issues, or request refactors — all without leaving your editor.
API keys are stored in VS Code's encrypted SecretStorage (OS keychain). They never leave your machine — we never see them.
Opus 4.6, Sonnet 4.6, Haiku 4.5, GPT-5.4 and more. Pick the right model for the task and switch anytime with Kazi: Configure.
Conversation history saved per workspace in .cody_memory.md. Context carries across sessions, refactors, and reboots.
Kazi Coder is available as a custom solution. Each installation is configured to your team's stack, preferred models, and internal tooling.
Schedule a DemoYou define the objective. Kazi's agent crew researches, plans, builds, and executes — with a human checkpoint before anything ships. Every step is tracked, auditable, and reversible.
Gathers context and surfaces the insights your workflow needs before any work begins.
Sets the brand voice, creative direction, and strategic framing before execution starts.
Produces the deliverables — drafts, reports, and actions — in the exact format you need.
Reviews every output for accuracy and alignment with your standards before it ships.
Review, edit, or reject. Nothing executes without your sign-off — ever.
Navigate, command, and control the entire platform by voice. Say "Hey Kazi" and your in-app AI assistant springs to life.
Say "go to workflows" or "show me the crew" — Kazi navigates the platform for you, hands-free.
Get spoken summaries of workflow status, agent activity, task progress, and quality scores in real-time.
Ask Kazi to launch a workflow, check task queues, or coordinate agents — all by voice command.
Kazi knows which page you're on and adapts responses. It highlights UI elements and can click buttons for you.
Voice agents, memory-backed chat, 100+ integrations, and background automation — everything you need to run intelligent workflows at scale.
Tell your agents what you need the same way you'd tell a teammate. Get status updates out loud, redirect mid-task, and stay in flow — no prompts, no copy-paste.
Agents send emails, update your CRM, post to Slack, and book meetings — without you touching a single app. Connect once, automate everything.
Know exactly what your agents did, what shipped, and what moved the needle. Every workflow run is logged — what executed, what was approved, and what the outcome was.
Your agents remember every client, deal, and conversation. No starting from scratch every session — they pick up exactly where you left off, every time.
Before anything sends, posts, or publishes — it lands in your queue first. Review it, tweak it, or kill it. Full output volume, full human control.
Agents run tasks around the clock — processing inputs, executing workflows, and handling the repeatable work — while you focus on decisions only you can make.
From 20-person startups to 500-person divisions — if your team runs repeated workflows, Kazi executes them.
Don't see your industry? Kazi runs any workflow your team repeats.
Talk to us about your use caseThe Kazi Agent Hub MCP server exposes the full execution layer — orchestration, token management, and real-time pipelines — directly inside Claude Code, your IDE, or any MCP-compatible client.
Invoke Riley, Jordan, Sam, Alex, and Cody directly from Claude or any MCP-compatible client. Full pipeline control from a single tool call.
Built-in token counting, context compression, and prompt caching — so your pipelines stay fast without blowing your budget.
Stream multi-step workflow results as they run. Create tasks, check status, and cancel operations — all from your preferred environment.
Python 3.11+ required · pip install -e . inside the folder · then add the config above to Claude Desktop
Scale your agent workflows to match your ambitions. Every plan includes the full execution layer — agents, orchestration, and approval queue.
Stop juggling tasks manually. Let your AI crew research, communicate, and quality-check every output — while you focus on what matters.
Launch Kazi Free
I built Kazi because I was tired of context-switching between a dozen tools just to get one intelligent thing done. Every agent has a job, every workflow has a checkpoint, and every action requires your approval before it ships. That's not AI hype — that's how it should work.