Open Source
For one pilot, self-hosted.
€0forever
- All core features
- Unlimited agents and flocks
- Approval rules and budget guard
- Community support on GitHub
Pricing
Free to fly. Pay when you scale.
flokpit is open source. Fly alone for free, or let us host the cockpit for your whole crew.
Team
Most popularWe host the cockpit. Fly together.
€29per pilot / month, billed yearly · €35 monthly
- Everything in Open Source
- Hosted control plane in the EU
- Co-pilots, on-duty rota and shift handover
- Pocket app, phone and email honks
- Autopilot, memory and flight recorder
- Integrations and 90-day audit log
Business
For bigger fleets and stricter rules.
€79per pilot / month, billed yearly · min. 5 pilots
- Everything in Team
- SSO, SCIM and two-vote approvals
- Org chart and org-wide checklists
- Unlimited audit log with export
- Self-hosted license option
- 99.9 % SLA and priority support
Enterprise. More than 50 pilots, air-gapped installs or custom contracts?
Let’s talk.
You pay per pilot, not per agent.
Fly 5 agents or 500. The price only changes when more humans join the cockpit.
AI models are billed separately.
Your agents use your own model accounts. Which ones you need.
How we set the price.
In line with tools teams already pay for: AI dev tools and agent platforms charge about €20 to €40 per seat for teams.
Proposed pricing for the concept. Prices excl. VAT and not final.
Compare
What’s in each plan
| Feature | Open Source | Team | Business |
|---|---|---|---|
| Agents, sub-agents and flocks | Unlimited | Unlimited | Unlimited |
| Formations, chat commands and honks | ● | ● | ● |
| Approval rules, budget guard, undo | ● | ● | ● |
| Flight plans, stuck detection, rest mode | ● | ● | ● |
| Bring your own AI accounts | ● | ● | ● |
| Browser notifications | ● | ● | ● |
| Who hosts the cockpit | You | We do (EU) | We do, or you with a license |
| Pilots | 1 | Per seat | Per seat, min. 5 |
| Co-pilots, on-duty rota, shift handover | – | ● | ● |
| flokpit pocket app (iOS, Android) | – | ● | ● |
| WhatsApp, Telegram, Signal and email honks | Self-hosted bot | ● | ● |
| Honk digest and quiet hours | – | ● | ● |
| Autopilot levels, flock memory, second opinion | Basic | ● | ● |
| Flight recorder and honk stats | – | ● | ● |
| Integrations (GitHub, Linear, Jira, Slack, Notion) | GitHub | ● | ● |
| Landing checklist | ● | ● | Enforced org-wide |
| Two-vote approvals | – | – | ● |
| Org chart and model compare | – | – | ● |
| Audit log | 7 days | 90 days | Unlimited, export and API |
| SSO (SAML, OIDC) and SCIM | – | – | ● |
| Uptime SLA | – | – | 99.9 % |
| Support | Community | Email, 2 business days | Priority, 4 hours |
Bring your own models
Your agents need model accounts.
flokpit is the cockpit. The thinking is done by AI models, and those are billed by the model providers, not by us. Connect at least one account. Each agent can use a different one.
A chat subscription is not an API account.
Plans like ChatGPT Plus, Claude Pro or Gemini Advanced are for chatting in the provider’s own app. Agents need an API key, usually billed per use (tokens). Set a spending limit at the provider and in flokpit’s budget guard.
Where to start
One API account is enough. Many teams start with one strong model for chief agents and a cheaper one for sub-agents. Or use OpenRouter to try several with one key. Keep data on your own machines with Ollama.
| Provider | Models | Good for in a flock | What you need | Billing |
|---|---|---|---|---|
| Anthropic | Claude | Coding, reviews, long tasks | API key from the Anthropic Console | Pay per use |
| OpenAI | GPT models | Coding, chat, tools | API key from the OpenAI Platform | Pay per use |
| Gemini | Large documents, multimodal | API key from Google AI Studio or Vertex AI | Pay per use, free tier with limits | |
| DeepSeek | DeepSeek models | Low-cost coding and reasoning | API key from the DeepSeek Platform | Pay per use |
| Mistral | Mistral models | EU-hosted option, fast small models | API key from La Plateforme | Pay per use |
| xAI | Grok | Chat and reasoning | API key from the xAI console | Pay per use |
| Cohere | Command models | Search, retrieval, enterprise text | API key from the Cohere dashboard | Pay per use, trial key |
| OpenRouter | Many models, one key | Trying and mixing models | One OpenRouter key and prepaid credits | Credits, per use |
| Amazon Bedrock | Claude, Llama, Mistral and more | Teams already on AWS | AWS account with model access enabled | Your AWS bill |
| Azure OpenAI | GPT models on Azure | Teams already on Azure | Azure subscription and a deployed model | Your Azure bill |
| Hugging Face | Open models | Open-source models, hosted | Access token, Inference Endpoints | Per hour or per use |
| Ollama | Local open models | Private data, no API cost | Your own GPU machine or cluster node | Free, you pay for hardware |
| Stability AI | Image models | Images for docs and posts | API key from the Stability platform | Credits |
| ElevenLabs | Voice models | Voice replies and audio | API key from ElevenLabs | Subscription with usage quota |
| Midjourney | Image models | Images | No official public API [check before launch] | Subscription |
Overview as of [DATE]. Model names, prices and terms change often. Check each provider’s website before you sign up. In the app: Settings → AI accounts.
Pricing FAQ
Questions about plans
Is the open-source version limited?
No. The open-source version has all core features. Paid plans add hosting, team features and support.
Do I pay per agent?
No. You pay for the cockpit, not the flock. Your agents run on your own infrastructure.
Do I need my own AI accounts?
Yes. Agents use API accounts from model providers like Anthropic, OpenAI or Google, billed by them. A chat subscription such as ChatGPT Plus is not enough. See the list.
Can I switch plans later?
Yes. Your flocks keep flying when you move between self-hosted and managed.