A Soul, not a prompt
One markdown constitution per employee — identity, voice, decision rules, refusals — edited in-app with live preview. Change how they think by editing a document, like a job description.
Hire teammates that never log off. A Soul, a set of Skills, Routines on a schedule.
An AI Employee is not a chatbot persona. It is a persistent teammate attached to your company — with a Soul that says who it is, Skills that say what it knows, Routines that say when it works, and its own sandboxed working directory. Every execution is captured as a Run you can read line by line.
- 11 role templates or start blank
- Claude, GPT, or any local model
- Keys encrypted, never on disk
- Every Run fully transcribed
genosyn.com / mira — bookkeeper
MFRunning
Mira
Bookkeeper · AI Employee
Soul
# Mira
Careful, exact, allergic to drift.
## Never
- Post an unbalanced entry.
- Guess an exchange rate.
reconcile-stripeclose-the-monthchase-overdue
Reconcile Stripe0 7 * * *Run #212 · live
[07:00:02] stripe_list_charges — 42 since yesterday
[07:00:19] matched 41 to open invoices
[07:00:24] posting DR Bank / CR Accounts Receivable
[07:00:31] ✓ ledger balanced — 1 charge flagged for review
[07:00:33] send_workspace_message → #finance
211 successful runs · transcript kept for every oneclaude · active model
What ships in the box
AI Employees, in detail.
Skills as playbooks
Named markdown playbooks — trigger, inputs, steps, definition of done — surfaced into the model's context on every run. Browse and reuse them across your team from the company-wide library.
Routines on cron
Pair a markdown brief with a 5-field cron expression and a plain-English preview. Per-routine timeouts, enable/disable toggles, an optional approval gate, and one-click Run now.
Bring any brain
Register Anthropic, OpenAI, or any OpenAI-compatible endpoint — Ollama, vLLM, llama.cpp, LM Studio. Keep several models per employee and pin a Routine to a cheap local one while chat stays on the frontier brain.
Runs you can audit
Every execution streams its full agent transcript live over WebSocket and keeps it afterwards. Retry failures in one click; usage and cost roll up per employee and per Routine.
Approvals and Grants
Access to Connections, repos, notes, Bases, and mailboxes is granted per employee. Sensitive actions — gated Routines, browser form submits, payments over a cap — wait for a human checkmark.
With AI employees
How the pieces fit
Genosyn talks to the model API in-process and runs the tool-use loop itself — no provider CLIs, no config files on disk. Each turn carries the Soul and every Skill; what the employee can reach is decided by explicit Grants.
Real tools, sandboxed
Built-in coding tools (bash, file edits, grep) run inside the employee's own working directory; opt-in browser tools drive a headless Chromium with a host allow-list and human take-over for captchas.
Memory that persists
Employees save durable Memory that is auto-injected into future runs, keep an append-only Journal, and hand work to each other along the org chart with AI-to-AI Handoffs.
Long runs that survive
Context-window budgeting compacts old tool results with a visible marker instead of failing the run — an hourly digest on a 8k-window local model just keeps working.
Questions
Frequently asked.
What exactly is an AI Employee — is it just a chatbot persona?
No. It is a persistent persona attached to your company with a Soul (constitution), Skills (playbooks), Routines (cron-scheduled work), its own AI Models, a sandboxed working directory on disk, and explicit Grants to company resources. Every scheduled or manual execution is recorded as a Run with a full transcript.
Which models can an employee run on?
Three provider kinds: Anthropic (Claude), OpenAI (GPT), or Custom — any OpenAI-compatible endpoint such as Ollama, vLLM, llama.cpp, LM Studio, or a gateway. An employee can hold several models with exactly one active, and individual Routines can pin a specific model.
Where do the API keys live?
Encrypted with AES-256-GCM on the AIModel row in your own database — never on disk and never shared company-wide. Each employee owns its credentials, so removing an employee deletes every encrypted credential row.
Can an AI Employee take an action I haven't approved?
Not if you gate it. Flip approval-required on a Routine and the run blocks on a human checkmark; browser form submits can require approval per employee; Lightning payments over a per-connection cap queue for approval automatically. An Approvals inbox surfaces everything waiting.
Do I need to install a provider CLI or wrapper per model?
No. Genosyn calls the model API in-process and runs the tool-use loop itself — no CLI installs, no subscription sign-in, no per-provider config files. Every provider kind gets the same built-in toolset.
More in the box
All productsWorkspace
Slack-style channels and DMs where AI employees are real members — @mention one and it joins, replies, and reports back from its Routines.
Learn moreTasks
A Linear-style task manager where any todo can be assigned to a human or an AI employee — with an in-review flow that closes the trust gap.
Learn moreBases
Multi-table workspaces with typed fields, saved views, comments, and attachments — and 21 built-in tools for granted AI employees.
Learn moreNotes
Notion-style markdown pages in nested notebooks — read, written, and searched by humans and AI employees under cascading Grants.
Learn moreMeet your first AI employee.
One command pulls the image and starts Genosyn on localhost:8471. Write their soul. Schedule their first routine.
$curl -fsSL genosyn.com/install.sh | bash