Pricing
You pay for what the models cost, plus a flat fee for the workspace.
The value metric is credits — one credit is a thousand tokens on a mid-tier model, one 1024px image, or one indexed document page. Every message in the app shows what it spent, so the bill is never a surprise at the end of the month.
Trying the thing before you trust it.
Free forever. No card, no trial clock.
Create a free account200 credits a month
- Chat with 6 open-weight models
- Standard and Temporary modes
- 20 images a month
- 3 documents, 40 pages each
- Provider named on every reply
- Private mode
- API access
One person who uses this every working day.
Billed monthly. Cancel any time.
Start Pro2,500 credits a month, then $0.008 each
- The whole catalogue, every vendor family
- All five privacy modes
- Image Studio: inpaint, upscale, variations
- Documents with page-level citations
- Research reports with tracked sources
- Characters with version history
- OpenAI-compatible API key
A group that needs one bill and an audit trail.
Billed monthly. Cancel any time.
Start a team trial8,000 credits per user, pooled
- Everything in Pro, per seat
- Shared characters and prompt library
- Org-wide privacy floor an admin can set
- Route log export
- SAML SSO and SCIM provisioning
- Role-based access, 4 roles
- Priority support, 1 business day
The prompt is not allowed to leave your network.
Support contract from $400/month, optional.
Read the deploy guideUnlimited seats, your hardware, your bill
- Source-available, AGPL-3.0
- docker compose up -d, five containers
- Bring your own provider keys
- Local model mode against your GPUs
- No licence fee, no seat count, no phone-home
- Support contract from $400/month
Prices in USD, excluding local taxes. Annual plans are billed once a year and refunded pro rata in the first 30 days.
The value metric
What a credit buys
Seat pricing punishes teams for adding people who barely use the thing. Pure pass-through pricing makes budgeting impossible. Credits sit in between, and the meter is visible.
| Action | Credits | Notes |
|---|---|---|
| 1,000 tokens · open-weight model | 1 | llama-4, qwen3, mistral-small |
| 1,000 tokens · mid-tier model | 3 | mistral-large-3, gpt-5.2-mini |
| 1,000 tokens · frontier model | 9 | claude-sonnet-4.6, gpt-5.2, gemini-3.0-pro |
| One 1024px image | 4 | 2K is 9, 4K is 20 |
| Inpaint a region | 3 | Only the masked area is regenerated |
| One indexed document page | 1 | Charged once, at ingest |
| One research report | 40–120 | Depends on how many pages it reads |
| Web search query | 2 | Standard mode only |
A realistic month
A security engineer using this as a daily driver — thirty or forty conversations a week, a few contract reviews, some diagrams — lands between 1,400 and 2,200 credits a month. That is what Pro is sized for. If you consistently need more, Team pools eight thousand per seat, which works out cheaper per credit.
Private mode is not upsold
Zero-retention routes cost providers more, so they cost you 12–18% more per token. We pass that through at cost and do not put a plan gate in front of it beyond the free tier, because charging a premium for the thing we say we believe in would be a poor look.
Every difference
Plan comparison
| Feature | Free | Pro | Team | Self-host |
|---|---|---|---|---|
| Models and routing | ||||
| Model routes available | Open-weight only | Whole catalogue | Whole catalogue | Yours, plus the catalogue |
| Provider named on every reply | ✓ | ✓ | ✓ | ✓ |
| Live price, context and latency in the picker | ✓ | ✓ | ✓ | ✓ |
| Switch model mid-thread | ✓ | ✓ | ✓ | ✓ |
| Fail-closed privacy enforcement | ✓ | ✓ | ✓ | ✓ |
| Privacy modes | ||||
| Standard | ✓ | ✓ | ✓ | ✓ |
| Temporary | ✓ | ✓ | ✓ | ✓ |
| Private (zero retention) | — | ✓ | ✓ | ✓ |
| Browser-only | — | ✓ | ✓ | ✓ |
| Local model | — | — | — | ✓ |
| Org-wide minimum mode | — | — | ✓ | ✓ |
| Modules | ||||
| Chat with streaming and markdown | ✓ | ✓ | ✓ | ✓ |
| Image Studio | 20 / month | ✓ | ✓ | ✓ |
| Inpaint, upscale, background removal | — | ✓ | ✓ | ✓ |
| Characters | 3 | Unlimited | Unlimited | Unlimited |
| Character version history | — | ✓ | ✓ | ✓ |
| Documents | 3 docs | 500 docs | 5,000 docs | Your disk |
| Research reports | 2 / month | ✓ | ✓ | ✓ |
| Team and platform | ||||
| OpenAI-compatible API | — | ✓ | ✓ | ✓ |
| Shared workspace | — | — | ✓ | ✓ |
| SAML SSO / SCIM | — | — | ✓ | Bring your own IdP |
| Route log export | — | ✓ | ✓ | Your database |
| Support | Community | Email, 3 days | Email, 1 day | From $400/month |
Scroll the table sideways to compare all four plans.
Pricing questions
Before you pick a plan
A thousand tokens on a mid-tier model, one 1024px image, or one indexed document page. Frontier models cost more credits per thousand tokens and the picker shows the multiplier before you send anything.
On Free, chat stops until the month rolls over. On Pro and Team, credits keep flowing at $0.008 each and the overage appears on the same invoice. You can set a hard cap per workspace if you would rather it stop.
On Team, up to one month's allocation rolls over and is pooled across the whole workspace. On Pro they do not roll over, which is why Pro is priced where it is.
The plan price is the same. The per-token cost is 12–18% higher on average because zero-retention routes are priced higher by the providers. The app shows the delta before you switch, not after.
Any time, from the billing screen, without an email to anyone. Downgrades take effect at the end of the period you have already paid for. Annual plans are refunded pro rata in the first 30 days.
The software is, under AGPL-3.0, with no seat count and no phone-home. You pay your own provider bills and your own hosting. A support contract with an SLA starts at $400 a month and is entirely optional.
Try it, then go and check the route log.
Free plan, 200 credits a month, no card. Every answer you get will tell you which company processed it — including the ones on the free plan.