Weave your optimal AI path
From real usage data, not guesses. WeavePrism turns your actual token flows into a living decision map — the right model for every task, the waste made visible, the savings one click away.
Choice paralysis is real
Hundreds of models, weekly price moves, benchmark noise. Deciding well takes hours you don't have — so most people don't decide at all.
Premium-by-default burns money
The classic failure: a frontier model quietly running your classification and summaries. It works — and it costs 10–50× what it should.
Generic advice isn't your advice
Leaderboards don't know your volumes, budgets or privacy lines. The right answer is a function of your usage — so it has to be computed from it.
Three passes through the prism
A genuine sequence: each step makes the next one smarter.
Shape your map
Pick your use cases and monthly volumes, drag five priority sliders, set hard constraints — budget caps, provider bans, open-weights only.
Weave in your usage
Drop in a usage CSV (or start with sample data). The map stops being generic advice and becomes a mirror of your actual stack.
Act on woven insights
The optimizer mines your last 30 days for savings, the simulator stress-tests each move, and one click pins it to your map.
Not another leaderboard. A loom.
Every feature exists to close the loop between what you pay and what you should pay.
Interactive Decision Map
Drag priorities, toggle constraints, watch the prism re-route every task to the model that earns it — live, not a static leaderboard.
Personal usage intelligence
Import CSVs from any provider. Spend by model, task and day; burn-rate forecasts; everything aggregated in your own database.
Proactive optimizer
“Move 65% of classification to a fast tier for ~42% savings, −2 quality points.” Concrete moves, ranked by dollars, grounded in your data.
What-if simulator
Test a migration on paper before you touch production: projected savings, quality delta and the new weave, instantly.
Proven Weaves
Fork community-validated setups from people with workloads like yours. Publish your own — votes and forks carry your name.
Privacy-first by design
Daily aggregates only, never raw prompts. One-click export and delete. Open-weights-only mode when data control is the priority.
Start from a Proven Weave
Curated and community-published configurations for workloads like yours — fork one and make it personal.
Free to weave. Pro when it pays for itself.
If WeavePrism doesn't find you more than $12/month in savings, don't upgrade. Simple.
Free
- Full interactive decision map
- 3 saved decision profiles
- CSV + manual tracking · 90 days
- Rule-based optimizer & simulator
- Fork & publish Proven Weaves
Pro
- Unlimited profiles & history
- Advanced what-if simulations
- Proactive savings alerts
- Export reports
- Priority catalog updates
- Founder pricing, forever
Team
- Everything in Pro
- Shared workspaces
- Usage policies & team analytics
- Admin controls
- SSO (roadmap)
Fair questions
How accurate are the recommendations?
Costs use official list pricing from a versioned catalog. Quality scores are benchmark-informed heuristics on a 0–100 scale — good for ranking trade-offs, not gospel. Every insight carries a confidence level, and we tell you to validate on your own evals before migrating traffic. The what-if simulator exists precisely so you can test before you commit.
What happens to my usage data?
CSVs are parsed in your browser; only daily aggregates (day × model × task totals) reach the database. No prompts, no completions, no API keys. Export everything as JSON or delete your account — and all of it — in one click.
Which providers are covered?
19 models across 10 providers today — Anthropic, OpenAI, Google, xAI, DeepSeek, Meta, Mistral, Qwen, Moonshot and Zhipu — including open-weights options you can self-host. The catalog is a single config file, so new models ship in minutes.
What's actually free?
The full decision map, CSV and manual tracking with 90 days of history, the rule-based optimizer, the simulator, community weaves, and 3 saved profiles. Pro adds unlimited profiles and history, deeper simulations and proactive alerts.
Do you take referral fees from model providers?
No. Neutrality is the product. Recommendations come from a deterministic engine over public pricing and open scoring you can inspect — nobody pays for placement.
Models change every week. Won't this go stale?
That churn is why WeavePrism exists. The catalog is versioned and data-driven; when prices or models move, one file updates and every map, insight and shared weave recomputes instantly.
Stop guessing which model deserves your tokens.
Three minutes to your first weave. Sample data included if you just want to see it think.
Weave your path