Platform plans from €0 to Enterprise: optimize tokens (local RAG, semantic cache, RAG context compression) and secure every prompt (PII detection, anonymization, sensitive-data governance). Free is a 30-day hosted trial; Starter and Growth use BYOK at provider rates.
Most popular — scale without compromise
€99 / month
PII, compression, and 4× quotas: the plan teams choose when LLM powers customer-facing apps in production.
Feature comparison — Growth
| Feature | Free€030-day trial · no credit card · hosted LLM Hosted LLMs, smart cache, and local RAG: test on real traffic and prove token savings before you subscribe. CacheRAG | Starter€29/ month BYOK, cache, and RAG in production: your provider keys and rates, with a lower LLM bill from week one. CacheRAG | Growth€99/ month PII, compression, and 4× quotas: the plan teams choose when LLM powers customer-facing apps in production. CacheRAGPIICompression | EnterpriseCustomAnnual & volume options Unlimited quotas, SLA, and a dedicated partner — for organizations where every token, model, and datum matters. CacheRAGPIICompression |
|---|---|---|---|---|
| Monthly platform fee | €0 | €29 | €99 | Custom |
| Requests / month | 500 | 50,000 | 200,000 | Unlimited |
| Requests / day50/day + 20 RPM fair use · monthly cap for platform trial | 50 | — | — | — |
| Rate limit (req / min) | 20 | — | — | — |
| Tokens / month | 150k | 10M | 100M | Unlimited |
| Tool-assisted responses / monthRuns with OpenAI tools enabled | 200 | 10,000 | 80,000 | Unlimited |
| TokenSaver API keys | 1 | 5 | 10 | Unlimited |
| RAG documents (quota)Upload limit when RAG module is enabled on your plan | 3 | 10 | 200 | Unlimited |
| Knowledge storage | 100 MB | 5 GB | 20 GB | Unlimited |
| Semantic compression | — | — | ✓ | ✓ |
| PII detection & anonymization | — | — | ✓ | ✓ |
| RAG retrieval moduleDocument upload & retrieval in pipeline | ✓ | ✓ | ✓ | ✓ |
| Upstream LLM (chat)Free: hosted entry models in quotas. Paid plans: BYOK — you pay providers at list price | 5 hosted entry models | 108 models BYOK | 196 models BYOK | 261+ full catalogue |
| RAG / cache embeddingsLocal open-source models (bge-large, etc.) — no OpenAI embedding API charge | ✓ | ✓ | ✓ | ✓ |
| Indicative model pricing (console)Pass-through estimate from llm_models catalogue; actual bill from your provider | Included (hosted) | Provider rates | Provider rates | Provider rates |
| Semantic cache (token savings)Hit rate and avoided tokens visible in console | ✓ | ✓ | ✓ | ✓ |
| OpenAI-compatible API | ✓ | ✓ | ✓ | ✓ |
| Console & usage monitoring | ✓ | ✓ | ✓ | ✓ |
| Self-service billing (Stripe) | — | ✓ | ✓ | — |
| Contractual SLA & dedicated support | — | — | — | ✓ |
| Sign up for free | Subscribe in app | Subscribe in app | Contact sales |
Quotas and modules match the TokenSaver console Plan & Usage view. Free is a 30-day trial with hosted entry-model inference within quotas. Starter and Growth are platform fees (cache, RAG, governance) — upstream chat LLM is BYOK at provider rates (see model pricing). RAG and semantic cache use local embeddings (included).
One governed pipeline — token savings and data compliance, visible in the console.
Local RAG on your documents · semantic cache (exact + similarity) · RAG context compression before the LLM · tokens saved per run in metrics
PII detection and anonymization (Growth+) · fewer leaks in outbound prompts · org governance, quotas, and traceability on every call
Measured in the console
Your knowledge base stays in your perimeter. The pipeline runs cache → RAG retrieval → compression of retrieved context → PII anonymization → LLM. Early adopter pilots see measurable savings on repeated prompts and document-heavy flows — tokens saved per run in the dashboard.
These are not marketing projections: they are TokenSaver modules measured in your dashboard.
When a similar request was already handled, the answer is served again without calling the LLM — you pay few or no tokens on that run.
Your documents stay in your perimeter; only useful excerpts go to the model, often compressed (Growth) to cut input tokens.
Every pipeline run shows billed vs saved tokens — use this to measure real ROI on your workflows.
Ballpark figures from early adopter pilots — your real numbers are in the console (tokens saved per run).
Subscribe in the TokenSaver console with Stripe Checkout (€29 or €99 per month). Your organisation plan and quotas update after payment. Change plan later via the Stripe Customer Portal.
A no-cost 30-day platform trial: hosted entry LLMs (no provider API key), semantic cache, and local RAG (up to 3 documents, 100 MB) — 500 requests and 150k tokens per calendar month during the trial, 50 requests per day, 20 req/min. Validate token savings and agent workflows, then upgrade to Starter or Growth (BYOK) to continue. Request access via the Early Adopter form — no credit card required.
Enable semantic cache, index documents in local RAG, and use RAG context compression (Growth) so the LLM only receives what matters. Each run shows billed vs saved tokens — ideal for support bots, copilots, and agents with recurring prompts.
The PII module (Growth and Enterprise) detects and anonymizes names, emails, and other entities before the provider call. Combine it with RAG and compression to limit business data exposure in prompts. Org-level quotas and traceability in the console.
When cache, RAG, and compression save more than the platform fee (€29 or €99/mo) — typical for repeated-prompt agents or internal-document flows. Growth adds PII and compression for compliance use cases. Real savings show up in the dashboard (tokens saved per run).
Free and Starter include cache and RAG. Growth and Enterprise add PII detection and semantic compression. Enterprise adds unlimited quotas, SLA, and dedicated support.
Free is a 30-day trial with hosted entry-model inference within plan quotas (no provider key to start). After the trial, upgrade to Starter or Growth. Starter and Growth charge a platform fee for cache, RAG, compression, PII, and governance — chat LLM is BYOK: you add provider keys and pay OpenAI, Anthropic, etc. at their official rates. TokenSaver monitoring shows indicative $/M from the model catalogue. RAG and semantic cache embeddings are computed locally (no embedding API bill). Enterprise is also BYOK with unlimited quotas and SLA.
Yes. Paid plans can be changed in-app via Stripe. Downgrading to Free may require cancelling your subscription in the portal first.
