Programmable LLM gateway. A thin substrate proxy in your AI request path. Auto-route, exact + semantic cache, prompt compression, batch arbitrage, cross-provider failover, per-stack quality canary — under a single drop-in SDK. Cost optimization is the wedge. Substrate is the endgame.
Free 60M tokens/month. Paid tiers are a flat monthly subscription by token volume — Starter $199, Growth $999, Scale $3,999, Enterprise custom. Keep 100% of measured savings.
→ tesseraai.io · tesseraai.io/dev · ledger.tesseraai.io
| Package | What it is | Install |
|---|---|---|
tessera-sdk |
Vanilla provider SDK patcher — wraps openai, anthropic, mistralai, groq, cohere clients. Start here. |
pip install tessera-llm-proxy · npm install @tessera-llm/tessera-sdk |
Drop-in adapters for the AI frameworks you already use:
| Framework | Package | Install |
|---|---|---|
| LangChain (Python + Node) | tessera-langchain |
pip install tessera-langchain · npm install @tessera-llm/langchain |
| Vercel AI SDK (Node) | tessera-vercel-ai |
npm install @tessera-llm/vercel-ai |
| LlamaIndex (Python) | tessera-llamaindex |
pip install tessera-llamaindex |
| Mastra (Node) | @tessera-llm/mastra |
npm install @tessera-llm/mastra |
| Pydantic AI (Python) | tessera-pydantic-ai |
pip install tessera-pydantic-ai |
| CrewAI (Python) | tessera-crewai |
pip install tessera-crewai |
| AutoGen 0.4+ (Python) | tessera-autogen |
pip install tessera-autogen |
All seven integrations route through the same proxy at api.tesseraai.io with a single tk_… key. Mix and match safely — they don't conflict.
import tessera
tessera.activate(api_key="tk_...")
# Your existing OpenAI / Anthropic / Mistral / Groq / Cohere code
# runs unchanged. Every request now flows through Tessera.That's it. No re-architecture. No new framework to learn. No code changes beyond activate(). Save measurements stream to your dashboard at ledger.tesseraai.io.
11 production mechanics, each gated by a per-workload quality canary that auto-disables on regression:
- M1 auto-route — cheaper-equivalent model swap when quality preservation holds
- M2 exact cache — request hash → cached response
- M3 prompt compress — lossless whitespace + structural compression
- M5 semantic cache — embedding-similarity lookup with cosine ≥ 0.95 default
- M6 prompt cache — Anthropic / OpenAI / Google native provider-side cache markers
- M7 context pruning — conversation history + RAG chunk trim past 12 turns or 32K chars
- M8 structured output —
response_format=json_schemainjection - M9 output-length predictor — max_tokens ceiling from p90 historical truncation rate
- M10 batch arbitrage — async-tolerant workloads routed onto provider batch APIs
- M11 cross-provider failover — opt-in OpenRouter fallback on 5xx / connection error
- C8 probe & prefill — foundation; per-ADR wire-in deferred
Per-stack quality canary auto-rollback fires per (workload, mechanic-stack) tuple — a regression on m1+m7 doesn't gate m1+m3. SLA floor 0.95 default; configurable per workload.
- Free Sandbox — 60M tokens/month, hard cap at quota, $0 fee, full mechanic stack active (route · cache · compress · batch). Solo devs + side projects covered entirely.
- Paid tiers — flat monthly subscription by gross tokens submitted: Starter $199 (≤1B), Growth $999 (≤5B), Scale $3,999 (≤20B), Enterprise custom (20B+). Full mechanic stack active, unlimited workloads + API keys, auto-rollback + 10% credit on quality regression. You keep 100% of measured savings.
No card up front. No trial expiry. No per-token markup, no cut of your savings — you keep 100% of what we save you.
- tesseraai.io — landing + product page
- tesseraai.io/dev — developer entry (Free Sandbox signup, install instructions, runnable examples)
- ledger.tesseraai.io — operator + sponsor dashboard
- tesseraai.io/blog — substrate-architecture deep-dives + cost case studies
- tesseraai.io/how-it-works — per-mechanic deep-dive trust track
- tesseraai.io/security — security posture, audit trail, SOC 2 progress
Tessera is operated by Fintechagency OÜ, Tallinn, Estonia (registry 16638667).