Evaluating Cabinet for your team? Book a demo
A free and open-source AI-first startup OS where everything lives as markdown files on disk. No database. No vendor lock-in.
No subscription. No trial. No paywall. Clone it, run it, and make it your own.
|
Why Cabinet
Search tools find. Chatbots answer. Cabinet does the work: on your files, your models, your infrastructure.
Everything is Markdown in a folder you own. Grep it, git it, back it up. No export, no lock-in.
Plug in the model accounts you already pay for: Claude, GPT, Gemini, Grok, local models. No markup, no new vendor.
Not just search and chat. A team of agents researches, drafts, and ships on a schedule, 24/7.
Testimonials
Cabinet is a big unlock for how I think about operating a business with AI.
Before Cabinet, AI felt fragmented: great for single tasks, but hard to orchestrate across the business. Every business has different workflows, and Cabinet makes it possible to turn those workflows into an AI team.
Florida, USA
Cabinet is exactly doing what it's strong at: orchestrating structure, standardization, opportunities, and challenges in a digital way.
Cabinet is the missing persistence and memory layer that TOGAF, ISO, and many other framework tools have never had. The framework provides the skeleton; Cabinet provides the living connective tissue.
Europe
As one of the first users of Superhuman, I definitely see the same spark here.
I use Cabinet to manage GTM and send emails through Apollo, and the whole system runs without me touching it, 24/7. The researcher agent built into Cabinet crafted award-winning GTM emails that blew my mind and achieved incredible open rates.
Customer stories
IllustrativeAnonymized scenarios that mirror how teams put an AI crew to work. Real, named case studies land here as teams go public.
reclaimed per rep each week
“Our reps stopped living in tabs. The SDR agent enriches and drafts overnight, so mornings start with reviewing, not researching.”
VP Sales · Series B SaaS
more publish-ready drafts per week
“Brand voice finally lives in a file instead of someone's head. New hires ship on-brand from day one, and the weekly recap writes itself.”
Head of Marketing · DevTools company
crown-jewel data never leaves
“It's the chief of staff I couldn't hire yet: it briefs me every morning and drafts the board update, all on infrastructure I control.”
Founder & CEO · Seed-stage startup
Bring your own AI
Cabinet runs on your existing model accounts and subscriptions. There's no bundled inference marked up on top, and no new AI vendor to push through procurement. Point it at what your team already uses.
…plus local models, and whatever comes next.
Integrations
Your company runs on dozens of tools. Cabinet brings the work scattered across them into one place you own, where your team and its agents can find it, act on it, and keep it moving.
…and the rest of the stack your team already uses.
Real Use Cases
Knowledge base + agents + files. One OS for wildly different workflows.
Solo Founder
“I run my entire startup from here. Strategy in /strategy/, roadmap in /product/roadmap.md, ICP in /market/icp.md. My CEO agent attends every planning session. I open a page, describe the week, and it challenges my assumptions and updates the mission board.”
Activity, last 24h
Agents
Files
Indie App Founder
“I use it to run my B2C app. The App Store listing is a markdown page I can update any time. I have an hourly Reddit scout that surfaces user complaints, and a weekly competitor job that crawls the market and dumps everything into /market/competitors/. I wake up to a briefing.”
Activity, last 24h
Agents
Files
Small Business Owner
“I have a CSV of leads. My agents read it, research each company, and draft personalised outreach. The pipeline CSV updates in real time. I watch it fill up from my dashboard. It's like having a sales team that never sleeps and never asks for a raise.”
Activity, last 24h
Agents
Files
Solo Creator
“Every Monday my Trend Scout scans HN and Reddit, picks the top signals, and my Draft Writer assembles the issue. I open Cabinet, read the draft, make a few edits, and hit send. What used to take 3 hours now takes 10 minutes.”
Activity, last 24h
Agents
Files
Freelance Consultant
“Each client gets their own folder. Meeting notes, proposals, deliverables: all markdown. When I switch clients, I just ask the AI about the folder. It has full context: the history, the decisions, the open questions. No re-reading, no catch-up.”
Activity, last 24h
Agents
Files
OSS Maintainer
“I linked my GitHub repo with .repo.yaml. When I merge a PR, an agent reads the diff, updates CHANGELOG.md, drafts release notes, and queues a Discord announcement. Maintenance overhead dropped to almost zero.”
Activity, last 24h
Agents
Files
Solo Founder
“I run my entire startup from here. Strategy in /strategy/, roadmap in /product/roadmap.md, ICP in /market/icp.md. My CEO agent attends every planning session. I open a page, describe the week, and it challenges my assumptions and updates the mission board.”
Activity, last 24h
Agents
Files
Indie App Founder
“I use it to run my B2C app. The App Store listing is a markdown page I can update any time. I have an hourly Reddit scout that surfaces user complaints, and a weekly competitor job that crawls the market and dumps everything into /market/competitors/. I wake up to a briefing.”
Activity, last 24h
Agents
Files
Principles
A few principles we think matter deeply for the future of AI + data tools. Every product decision gets weighed against these.
Own and export your data, anytime
Your knowledge base is plain markdown on disk, not rows in a vendor's database. Run ls and you see your work; move the folder and the whole cabinet moves with it. There is no export request to file, no proprietary format to decode, and no one sitting between your team and its own files.
Yours=you can take everything and leave whenever you want.
Every change tracked and reversible
Every save auto-commits, so your team can inspect diffs, review how knowledge evolved, and restore any page to an earlier state. You get the same accountability over your AI's output that you already expect from production code.
Git everything=nothing is ever silently lost or overwritten.
Use the AI you already pay for
Cabinet connects to the model accounts your organization already holds: Claude, Codex, OpenCode, local models, and whatever ships next. You bring the inference, so spend flows through contracts you control rather than a markup layered on top.
BYOAI=no second subscription and no per-seat AI markup.
Simple enough to audit
Plain files and predictable behavior mean there is no black box to take on faith. The system is small enough for your engineers to reason about and verify themselves. If a capability can't be explained in a paragraph, it usually shouldn't ship.
KISS=your engineers can verify exactly what it does.
Human approval before anything leaves
Every dispatched task waits in a human approval queue before it sends an email, calls an API, or touches anything external. Exposure is limited by default, so giving agents autonomy never means giving up control.
Security=an agent never acts on the outside world unsupervised.
Runs in infrastructure you control
Cabinet is open source and self-hosted by default, so plans, research, and operating memory live where your existing policies already apply. The only data that ever leaves is the prompt and response of a cloud-model call you explicitly invoke.
Self-hosted=your context stays inside your own environment.
If a feature would break any of these, it doesn't make it in.
Install Flow
The install walkthrough is back. One command scaffolds the workspace, sets up your AI team, and gets Cabinet running locally.
The Problem
Every time you start a new Claude session, it forgets everything. Your project context, your decisions, your research. Gone. You keep re-explaining the same things. Cabinet gives your AI a persistent brain: a knowledge base that both you and your agents read and write to, 24/7.
Scattered docs in Notion. AI sessions that forget context. Manual copy-paste between tools. Scripts held together with tape.
One knowledge base. AI agents that remember everything. Scheduled jobs that compound. Your team grows while you sleep.
If it feels like enterprise workflow software, it's wrong. If it feels like watching a team work, it's right.
Unique to Cabinet
This is the biggest difference between Cabinet and tools like Obsidian or Notion. Drop an index.html into any folder and it renders as a live, interactive app.
.app marker. Sidebar and AI panel auto-collapseFeatures
A complete operating system for your startup: knowledge base, AI agents, task management, and team chat in one self-hosted app.
Onboard a CEO, Editor, Marketer. Each has goals, skills, scheduled jobs. Watch them work like a real team.
No database. Markdown on disk. Drag-and-drop tree sidebar. Your data is always yours, always portable.
Rich text editing with Tiptap. Tables, code blocks, slash commands. Toggle to raw markdown anytime.
Drop an index.html in any folder. It renders as an iframe. Full-screen mode for dashboards.
Full Claude Code terminal in the browser. xterm.js + node-pty. Run commands without leaving Cabinet.
Every save auto-commits. Full diff viewer. Restore any page to any point in time. Linked repo support.
Cron-based agent automation. Reddit scout every 6 hours. Weekly reports on Monday. Your AI team never sleeps.
Break goals into missions. Assign tasks to agents. Track progress with Kanban boards.
Built-in team channels. Agents and humans communicate. @mention an agent to trigger a response.
Cmd+K instant search across all pages. Fuzzy matching. FlexSearch index rebuilt on every change.
PDFs render inline. CSVs open as editable tables with add/delete rows and columns.
Add .repo.yaml to link KB directories to source code repos. Agents read your codebase.
AI Team
Answer 5 questions. A CEO agent appears. It suggests teammates. Each agent has goals, skills, and recurring jobs. You watch them work like watching a real team.
CEO Agent
Lead
Strategic planning, goal tracking, task delegation. Creates missions, coordinates the team.
Editor
Specialist
KB content editing, formatting, linking. Maintains documentation quality.
Content Marketer
Specialist
Blog posts, social media, newsletters. SEO-optimized content generation.
SEO Specialist
Specialist
Keyword research, site optimization, ranking analysis.
Sales Agent
Specialist
Lead generation, outreach, pipeline tracking.
QA Agent
Specialist
Review, proofread, fact-check content across the KB.
How It Works
One command. Next.js + daemon start. Your knowledge base is a /data directory on disk.
npx cabinetai runWhat's your company? What do you do? What are your goals? Cabinet builds your custom AI team.
Agents create missions, write content, scout Reddit, review quality, all on schedule.
Every agent run, every edit, every research session adds to the KB. Your system gets smarter every day.
Solutions
Same knowledge base: a different AI crew for each function, working on your data, on your infrastructure, around the clock.
Pipeline research, outreach, and CRM hygiene on autopilot.
Explore
Always-on research, content drafts, and campaign reporting.
Explore
Living docs, incident write-ups, and codebase-aware agents.
Explore
Synthesized research, living PRDs, and competitive tracking.
Explore
SOPs, recurring reports, and cross-tool reconciliation.
Explore
An always-on chief of staff across the whole business.
ExploreThe business case
The numbers execs actually care about, before a single line of custom integration.
per-seat AI markup: you bring your own models
of your data stays on your infrastructure
replaces the wiki, the chatbot, and the agent platform
your AI team runs on a schedule, not on demand
Built for organizations
The people who sign off on new tools (security, legal, platform owners) should come away with less to worry about, not more.
In progressWe're working toward it, and we'll show you the report, not just a badge.
MIT licensed. Read every line, fork it, or run your own build.
Runs in your environment, under the controls you already enforce.
Your prompts and content are never used to train a model.
Your model providers, your API keys. Inference never routes through us.
Every change to data and agents is versioned, diffable, and attributable.
Enterprise trackSAML single sign-on and provisioning for teams that need it.
Self-hosted means your knowledge lives in your region, under your policies.
Questions for a vendor review? hi@runcabinet.com
Cabinet is a free, open-source project you can run yourself. No subscription, no trial clock, and no vendor lock-in. Start in 2 minutes.
Questions? hi@runcabinet.com
Hosted version coming soon. No setup, automatic updates.
First 500 early supporters get priority access.
Hosted version coming soon
Get launch updates and early access when Cabinet Cloud is ready.
Get launch updates by email. Unsubscribe anytime.