# Linzumi

> A team chat platform that lets you command a fleet of coding agents running on your own machine, with full visibility and control from any device.

Linzumi is a team chat platform built around coding agents that run on your own hardware — not in a hosted cloud sandbox. Currently in beta and available as a macOS app, it lets developers steer multiple AI coding agents in parallel from any device, including a phone, while teammates watch diffs land and redirect runs mid-task inside the same thread. The company is backed by Y Combinator, Matrix, SV Angel, Decibel, Pioneer Fund, and Axiom Partners.

## What It Is

Linzumi sits at the intersection of team messaging and AI coding agent orchestration. Instead of running agents in a remote VM or leaving them unattended in a terminal, Linzumi keeps agents on the developer's own machine while surfacing every action — commands run, files touched, diffs produced, test output — directly inside a shared chat thread. The result is a single channel where humans steer and agents build, with the whole team able to jump in mid-run.

## How the Agent Fleet Works

Each coding job runs on the operator's local machine using their own credentials and API keys. The agent posts diffs, commands, and test results back into the thread so teammates can review inline without leaving chat. Key mechanics the homepage describes:

- **Parallel steering**: One operator can run multiple agents simultaneously; multiple teammates can each steer their own agents in the same workspace.
- **Mid-run intervention**: Teammates can redirect an agent before it solidifies a bad approach, catching problems when they are cheap to fix.
- **Remote control without SSH**: Agents are steerable from a phone or any browser — no SSH or terminal app required.
- **Live preview**: Build output is visible in the thread before the operator returns to their desk.

## Trust and Control Model

Linzumi's homepage describes a layered permission system designed to keep agents inside explicit guardrails:

- Agents only access directories the operator explicitly approves; the rest of the machine is invisible.
- Forwarded ports are scoped to the active thread and close when the conversation ends.
- Permissions expire with the session and must be re-granted explicitly.
- Network egress, credential reads, and repo writes require explicit operator approval — agents cannot escalate themselves.
- Every action (commands, files, screenshots, test output) is attached to the thread and exportable for audit.
- A dedicated `#linzumi-support` channel is created in every workspace; the Linzumi team can read it but cannot access other threads, repos, tokens, or Codex transcripts.

## Roadmap: C3 Continuous Context Compiler

The homepage lists a coming-soon feature called C3 (Continuous Context Compiler), described as a system that turns every chat message, call, and coding job into queryable, cited knowledge. Planned capabilities include live call transcription (by joining a meeting link), contradiction reconciliation across channels, and automatic generation of verifiable specs that drive coding work. This feature is not yet available.

## Current Status: Beta

Linzumi is currently in public beta. The macOS desktop app is available for download, and a web sign-up path also exists. The homepage describes the beta as free with no credit card required for the personal tier. The product is actively being developed, with C3 listed as an upcoming roadmap item.

## Features
- Fleet of coding agents running on your own machine
- Team chat with inline diff review
- Mid-run agent redirection from any device
- Parallel multi-agent steering
- Directory-level ACLs and permissioned runtime
- Full audit log of every agent action
- Forwarded ports scoped to thread lifetime
- Live build preview in chat
- Run replay and full run logging
- Dedicated #linzumi-support channel in every workspace
- Upcoming C3 Continuous Context Compiler
- Voice, phone, and on-call workflows (Company plan)
- Self-hosting and VPC options (Enterprise)
- SSO/SAML and SCIM provisioning (Enterprise)
- No per-seat pricing on team plan

## Integrations
OpenAI Codex, macOS, Your own API keys and credentials

## Platforms
MACOS, IOS, WEB, API

## Pricing
Freemium — Free tier available with paid upgrades

## Links
- Website: https://linzumi.com
- EveryDev.ai: https://www.everydev.ai/tools/linzumi