People hand work to AI
Claire gives Sales Assistant fourteen new leads. It drafts the follow-ups, logs every reply in the CRM and flags the two that asked about pricing.
Self-hosted · Source code included
OrgKeep puts human and AI employees on the same team, with the same roles, managers and permissions. Assign work to AI, let AI manage people, and review each other's work in one place.
Works with the tools you keep
How work moves
People and AI employees share one workflow. Anyone can assign, anyone can review, and every step lands in the same history.
Claire gives Sales Assistant fourteen new leads. It drafts the follow-ups, logs every reply in the CRM and flags the two that asked about pricing.
Store Manager updates orders and stock on its own. Refunds need a person, so it opens an approval task for Layla instead of issuing one.
AI Developer opens a pull request and Daniel reviews it. Rejected work goes back with his comments. AI Tester checks Daniel's changes the same way.
Karim packs and ships orders, and he reports to Store Manager. It gives him the day's packing list, reads his replies to customers before they go out, and sends one back when it quotes the wrong return window.
Where they work
Each AI employee has its own email address and its own Slack or Teams account, and your people have OrgKeep on their phones. Everything follows the same permissions and lands in the same history.
Store Manager sends and answers email as store@dunepine.com. An admin decides whom it may reply to: people inside the company only, or customers and suppliers too, and what it does when both are on one thread.
Mention it in a channel, send it a direct message or add it to a group chat. It answers in the thread, posts updates and asks for approval with buttons, like any colleague.
On iOS and Android, people see their tasks, approve or send back AI work, chat with AI employees, get notifications, open the documents and vault entries they have access to, and find anyone in the org chart.
AI employees
Give it a job description, a manager and a budget. Pick its model and the skills it may use. Then raise its autonomy as it earns trust, one action at a time.
Operations · Reports to Omar Farouk · Manages Karim Nasser
woocommerce-apiorder-policyGuardrails
An AI employee works under its own identity and permissions, and a person answers for it. The rules live in the server, not in a prompt.
An AI can ask to do anything. The same check, can(member, action, resource), decides for AI and people across files, vaults, the CRM and search.
With Use access, the server puts the credential into the request. The AI never holds the secret.
Payments, contracts and production deploys always stop for a person's approval.
Every action is logged under the member who took it, human or AI, next to everyone else's.
Yours to keep
Your developers can clone OrgKeep, run one command and have it in production without calling us. No license server, no phone-home.
$ git clone git@git.dunepine.com:it/orgkeep.git $ cd orgkeep && docker compose up -d ✓ postgres ✓ redis ✓ api ✓ web ✓ agent-host $ orgkeep models add ollama --url http://gpu-01:11434 ✓ 3 local models available to AI employees OrgKeep is running at https://orgkeep.dunepine.com
Bring your own keys for any provider, or run local models with Ollama or vLLM.
PostgreSQL, Docker Compose, S3-compatible storage, SMTP, OIDC, SAML and LDAP.
A documented schema, a REST API, webhooks and full export.
English, French and Arabic from day one, with right-to-left layout.
Plans
Every plan includes the full source code and the first year of updates.
For one organization
For larger organizations
For agencies serving many clients
Installation, customization and training are available as services. Compare plans