ExecBot gives you a personal crew of AI specialists that share your context, learn from your history, and work together under one chief of staff — Atlas. Built for people who want to stop pasting the same briefing into ChatGPT ten times a week.
Most AI tools ship with generic personas. ExecBot builds a crew from your actual work patterns — then lets you edit every prompt before anything goes live.
Sign in with Google. Paste your Anthropic API key. It's encrypted with AES-256-GCM before it touches disk.
Drop a ChatGPT or Claude export. Atlas reads hundreds of your past prompts to understand what you actually work on.
Atlas proposes 3-7 specialists tailored to your domains. Edit their names, prompts, and icons. Approve the ones you want.
Talk to Atlas. It routes to the right specialist, collects results, and hands you a synthesized answer.
Not a chatbot, not a playground. A real workspace with memory, delegation, swarms, automation, and observability.
Atlas reads every task, knows every specialist, and routes work for you. Stop context-switching between tabs.
Create agents with precise prompts, icons, and schedules. Every agent sees the same shared task board.
Pre-built teams that work as one — a finance desk led by a CFO, a marketing department led by a CMO. Give them a brief and they work it in phases: audit, strategy, execution, each step building on the last. Exec Elite
Drop your ChatGPT export. Atlas proposes a crew tailored to what you actually work on — not generic templates.
Your Anthropic API key, encrypted at rest. Test it with one click, revoke anytime, no vendor lock-in — and watch every dollar in a live spend meter with a daily cap you set.
Your agents, tasks, and delegations as a force-directed 3D graph. Drag to rotate, click to dive in. Powered by Three.js.
ElevenLabs ConvAI voice mode, on web and mobile. Tap the mic, speak, and Atlas listens, delegates, and answers out loud — no typing required. Exec Pro + Voice
Set a topic, set a cron, get briefed. Atlas runs your crew on a recurring schedule and delivers the synthesis. Morning briefings without lifting a finger.
Triggers that fire without you. When an email matching your rule arrives, Atlas runs your crew and drafts the response — held behind a send-permission gate so nothing leaves your outbox until you approve it.
Keep separate worlds separate. Scope conversations, crews, briefings, and automations to a project — Atlas stays aware of which one it's working in.
Atlas and the crew remember. Past decisions, preferences, and context carry across sessions — no re-briefing, and the more you use it the sharper it gets.
Your key, your ceiling. Set a daily spend cap and watch every request at provider cost in a live meter — total transparency, zero markup.
Every tier is bring-your-own-key — your usage, your cap, zero markup. Annual billing saves two months.
ExecBot is a multi-tenant SaaS with row-level security at the database layer. Your data never mixes with anyone else's, and your keys never leave your tenant boundary.
API keys encrypted with a separate tenant key-encryption key. Plaintext keys are never logged, never cached, never returned through the API.
Postgres row-level security enforces tenant isolation. Every query is scoped to your tenant ID at the connection level, not the application layer.
Your conversations are never used to train any model. Anthropic's API doesn't train on customer data. ExecBot adds nothing on top of that.
Yes. ExecBot is BYOK — you paste your Anthropic key once, it's encrypted, and it's used for all agent calls in your tenant. We don't resell tokens. Bonus: you see every request cost at the provider level.
Samples are sent to Claude once for analysis (to propose a crew). Nothing is stored persistently — the proposed crew is written to your tenant, but the raw samples are dropped after analysis. You can also paste text directly instead of uploading a file.
Yes. Every agent — seed crews, imported agents, custom agents — has a fully editable prompt, icon, color, and description. Hover a row, click the edit icon.
Three ways: (1) they all share one task board so Atlas can delegate cleanly, (2) they can run on schedules without you opening the app, and (3) they see each other's outputs for synthesis.
A ready-made team of specialists with an orchestrator at the head — a CFO-led finance desk, a CMO-led marketing department. Hand it a brief and it works the problem in phases, then returns one synthesized result. You can build your own, too.
Within limits you set. Briefings run your crew on a schedule; event triggers fire Atlas off an incoming email that matches your rule. Anything that sends on your behalf waits behind a send-permission gate until you approve it — automation never goes rogue.
Not yet. ExecBot is a hosted product for now. If that changes, the core agent engine is the piece most likely to go first.
One person. Built in public across several long sessions. See the roadmap in the app for what's shipped and what's next.