Choose where your prompts live. Pay for the platform, not your tokens.
Open-source data plane. Cloud console free for 30 days. Your own LLM keys on every plan (BYOK), with zero markup: you pay your providers directly, at their prices. Paid plans are a flat fee per organisation.
1 to 3 edge gateways · proposal within 24 business hours
Your prompts:your infrastructure
Your gateway at home. Your control plane in the cloud.
LLM traffic stays on your infrastructure; console, policies and aggregated FinOps stay in our cloud. Up to 3 edges on the cloud control plane, no Enterprise contract.
1 edge gateway included, up to 3 on the same control plane
BYOK on the edge: prompts, keys and detailed replay stay there
Paid plans cover the console, governance and, depending on the plan, the hosted gateway or your edges.
Your LLM, your keys
On every plan, Free included, you bring your own keys: you pay OpenAI, Anthropic and others directly, at their prices.
No cut of your tokens, no seat tax
TokenSaver never resells tokens or takes a cut of them. Paid plans are a flat fee per organisation, not per user.
Pro Hybrid, Enterprise or a specific need
A proposal within 24 business hours.
Pro Hybrid, Enterprise, self-host or a question about the plans? Four clicks and your contact details: we come back with a proposal sized to your deployment and your volumes.
✓An answer from the founding team, not a bot
✓No markup on your tokens: you only pay for the platform
Console, AI flows and modules on your own LLM keys. TokenSaver includes no tokens: you see your real provider costs from day one.
After day 30
Only the control plane and the cloud console stop. A gateway running on your side, enrolled or not, keeps running 100% autonomously: you configure it by hand, with no quota limit, and it keeps serving with your keys.
Keep the console
Pro SaaS (€39 / month) if the gateway runs with us, Pro Hybrid (on quote) if it runs with you, Enterprise for the security committee.
Quotas
Free
Pro SaaS
Pro Hybrid
Enterprise
Requests / month
50,000
100,000
No limit (your edge)
500,000
Tokens / month
Not capped (BYOK)
Not capped (BYOK)
Not capped (BYOK)
Not capped (BYOK)
Requests / day
5,000
10,000
No limit (your edge)
20,000
Requests / minute
120
200
No limit (your edge)
300
Responses with tools / month
10,000
Unlimited
Unlimited
Unlimited
Tool definitions per request
256
256
256
512
TokenSaver API keys
1
20
20
100
RAG documents
10
2,000
2,000
10,000
Knowledge storage
200 MB
10 GB
10 GB
50 GB
Edge gateways
—
—
1 included, up to 3
More than 3
Content sent to the console (hybrid)
—
—
4 GB, kept 30 days
By contract
Beyond the included volume
—
€9 per 100,000 requests, pro rata
No overage: traffic runs on your edge
By contract
Tokens are never capped: you use your own keys on every plan. Free quotas guard the platform during the 30-day trial; they are not LLM credit, and a gateway running on your side after the trial has no quota limit. Pro Hybrid keeps the console quotas of Pro SaaS (keys, RAG, storage); its requests go through your own edge, with no TokenSaver quota limit. Enterprise values are a starting point, set by contract.
What’s included
Free
Pro SaaS
Pro Hybrid
Enterprise
Deployment and keys
Data plane
Cloud during the trial, or local once released
TokenSaver cloud
Your edges (1 to 3)
Edges, on-premise, air-gap
Cloud console
30 days
By contract
Prompts and detailed runs
SaaS gateway (TokenSaver cloud) or open-source gateway (your infrastructure)
TokenSaver
Your infrastructure
Your infrastructure or dedicated instance
Your provider keys (BYOK)
No markup on tokens
Observability and IDE
Execution graph and cost per Run
Open-source VS Code / Cursor extension (once released)
IDE policies (PII, cache, compression)
30 days in the cloud, no limit on a local gateway
Optimisation and modules
Semantic cache
Local RAG
Compression
PII detection & anonymisation
Agentic loops
Governance, security and support
Agent Registry (catalogue, approval)
Smart routing
Self-serve checkout (Stripe)
CCR (recoverable compressed originals)
Cyber pack (policies, posture, alerts, audit)
SSO (OIDC) + SCIM
Chargeback and cost centers
Execution graph scoped by team and project
SIEM audit export
Team memory, agent traces, context graph
SLA, dedicated support, on-premise
The Enterprise cyber pack coversPrompt injectionDecoratorsTool allowlistOutput handlingContent moderationModel routing
Frequently asked questions
Are LLM costs included?+
No, on any plan. You use your own provider keys (BYOK) and pay OpenAI, Anthropic and others directly, at their list prices. TokenSaver includes no tokens and takes no markup.
Pro SaaS or Pro Hybrid?+
It comes down to where your prompts live. Pro SaaS: the gateway runs on our cloud, nothing to operate. Pro Hybrid: the gateway runs on your infrastructure and prompts, keys and detailed replay stay there, while the console stays in our cloud. Pro SaaS is €39 per organisation per month; Pro Hybrid, with 1 to 3 edges, is on quote.
What does the control plane receive in hybrid mode?+
Metadata for AI flows and aggregated FinOps, plus gateway health signals. Prompt content only goes up if you opt in. Your prompts do go to the LLM provider you choose, unless you run on-premise models under Enterprise.
Is the gateway open source?+
The edge gateway will be released as open source; the public repository is not available yet. The control plane is not open source, and Enterprise features are offered under a commercial license.
Can I keep my Claude subscription?+
Yes. With the open-source VS Code and Cursor extension, coming soon to the VS Code Marketplace, Keep my Claude mode leaves Claude Code on your Anthropic plan: no LLM API key, no extra LLM bill. On Free, cloud policies for PII, cache and compression run for 30 days. After that, Keep my Claude keeps working: you plug it into a local gateway, the open-source data plane, with no quota limit. Paid plans keep cloud policies on with no limit.
What happens after the 30-day trial?+
Only the control plane and the cloud console stop. A gateway running on your side, enrolled or not, keeps running 100% autonomously: you configure it by hand, with no quota limit, and it keeps serving with your keys. The free VS Code extension, Keep my Claude mode included, plugs into it locally. Move to Pro SaaS, Pro Hybrid or Enterprise to get the console and centralised control back.
Pro Hybrid or Enterprise?+
Pro Hybrid covers up to 3 edges on the cloud control plane, on quote. Enterprise is for more than 3 edges, SSO, the cyber pack, air-gap and a contract with an SLA.
How do I move to a paid plan?+
Create an account, then upgrade to Pro SaaS in the console, by card through Stripe, billed monthly per organisation, excl. VAT. Beyond the 100,000 included requests, Pro SaaS keeps serving and adds €9 per 100,000 requests, pro rata, to the next invoice; you get alerts before you reach the limit. Pro Hybrid and Enterprise go through a quote: use the form or contact sales.
What does Enterprise add?+
SSO and SCIM, the cyber pack, CCR, chargeback, audit export, a graph scoped by team and project, more than 3 edges, air-gap, an SLA and dedicated support, under an annual commercial license.
What is CCR?+
CCR keeps the originals behind compression and lets agents retrieve them on demand, for audit, compliance and regulated sectors. Enterprise only.
Where is my data hosted?+
It depends on the mode. In SaaS, on the TokenSaver platform, in European datacenters. In hybrid, prompts and keys stay on your edge and the console receives metadata. In self-host, everything stays with you.
Early Adopter Program
Building something bigger? Let’s talk.
A personalised demo, a governed 30-day POC and Early Adopter terms for teams who want to shape TokenSaver with us.
