mintoken
Start free →
Operations

Changelog

Platform updates, new endpoints, and breaking changes — most recent first. Older entries trimmed to high-impact only; full history lives in the public GitHub repo.

2026-05-02
v0.7
16 providers, per-key savings, dashboard polish.
  • Provider registry expanded from 10 to 16: added Fireworks AI, Cerebras, SambaNova, NVIDIA NIM, Hyperbolic, and DeepInfra alongside the existing OpenAI, Anthropic, Google, DeepSeek, Mistral, xAI/Grok, Groq, Together AI, Perplexity, and OpenRouter. All available in the playground and via GET /v1/playground/providers.
  • Dashboard now shows per-key savings on the API keys page — see exactly which key is doing the work, and a 14-day savings sparkline on the overview so trend at a glance.
  • New per-model breakdown in analytics, plus a public GET /v1/stats/public endpoint surfacing aggregate tokens-saved.
  • First-run onboarding panel on the dashboard walks new users through key creation, plus a self-serve forgot-password flow and a polished settings page.
  • Login Suspense boundary fix unblocks production builds; duplicate API-key names now rejected with a clear error.
2026-05-03
v0.6
Context compression — Tier 1 ships.
  • New: opt-in context_compress per API key. Compresses the messages[] array before forwarding to the upstream provider. Targets the cache token reads that dominate the bill in long agent sessions.
  • Tier 1 techniques: tool-result truncation (head + tail with a clear marker) and cross-turn deduplication (same file across multiple turns becomes a tiny back-reference). Pure heuristic — no ML, no extra LLM calls, zero added latency.
  • Three levels: light / standard / aggressive. Recency window of 4-8 turns is always preserved verbatim. Code blocks, error messages, file paths, and in-flight tool_use blocks are never touched.
  • Response headers when active: X-Mintoken-Context-Tokens-Saved / X-Mintoken-Context-Tokens-Before / X-Mintoken-Context-Tokens-After. Same numbers tracked in usage records and surfaced on the dashboard.
  • New docs page: /docs/context-compression. Full landing section explains the 1.5B-token / 95%-cache problem.
  • Migration 003_context_compress.sql — adds context_compress + context_compress_level to api_keys and three tracking columns to usage_records.
2026-04-19
v0.5
10-provider playground + BYO endpoints.
  • Playground now supports OpenAI, Anthropic, Google, DeepSeek, Mistral, xAI/Grok, Groq, Together AI, Perplexity, and OpenRouter out of the box.
  • New Custom (BYO endpoint) provider — point mintoken at any OpenAI-compatible URL (Ollama, vLLM, LiteLLM, internal proxies).
  • New public endpoint GET /v1/playground/providers — returns the live provider + model registry. Use it in your own UIs.
  • Login now reports specific Supabase errors (e.g. "email not confirmed") instead of always falling back to "invalid credentials".
2026-04-18
v0.4
Razorpay billing live + search-engine readiness.
  • Self-serve plan upgrades on the dashboard. Pro ₹1,599/mo, Team ₹3,999/mo. UPI, cards, net banking, wallets, EMI all supported via Razorpay Checkout. GST invoices auto-emailed.
  • New endpoints: POST /v1/billing/razorpay/order, POST /v1/billing/razorpay/verify, POST /v1/billing/razorpay/webhook.
  • SEO pack: dynamic sitemap at /sitemap.xml, generated 1200×630 OpenGraph image, JSON-LD structured data (Organization + WebSite + SoftwareApplication + FAQPage + BreadcrumbList) — all docs and landing pages eligible for Google rich results.
  • New logo mark — rounded ink "token stamp" with cream + orange bars. Replaces the descending-bars Flow mark.
  • Fixed: clicking the logo from inside the dashboard now lands you on the dashboard, not the marketing site.
2026-04-17
v0.3
Dashboard, playground, full docs.
  • Dashboard pages live: overview, API keys (create / rotate / revoke), playground, billing, settings.
  • Playground compares your prompt with and without mintoken side-by-side — see token savings, copy responses, share results.
  • Full docs tree: quickstart, authentication, OpenAI / Anthropic / Google adapters, intensity, custom rules, smart-detect, streaming, rate limits, errors, analytics.
  • Marketing site shipped: hero, savings calculator, benchmarks, pricing, FAQ, providers marquee.
  • Production wired end-to-end: frontend on Vercel (mintoken.dev), backend on Render (mintoken-api-36nc.onrender.com), Supabase auth (HS256 + ES256/RS256).
2026-04-16
v0.2
Multi-provider compression backend.
  • /v1/compress + /v1/optimize can use either OpenAI or Anthropic as their internal compression LLM, configured per deploy.
  • Plan-aware quotas: free / pro / team / enterprise tiers with per-plan token, key, RPM, org-member, and rule-count limits.
  • Smart auto-detect — picks the right intensity based on prompt content (security / legal / code / explanation).
2026-04-14
v0.1
Production proxy launch.
  • OpenAI (/v1/chat/completions), Anthropic (/v1/messages), and Google (/v1/generateContent) drop-in proxies live.
  • Three intensity levels: lite / full / ultra.
  • Streaming passthrough for OpenAI + Anthropic.
  • Analytics API: summary, time-series, per-key, per-model breakdowns.
  • Free tier: 100,000 tokens / month, 2 API keys, no card required.
Coming soon
Roadmap
  • Razorpay Subscriptions — real auto-debit recurring billing (currently one-time orders; users re-pay each month).
  • Stored provider keys — save your OpenAI / Anthropic / Google key in your account (encrypted at rest) and skip the X-Provider-Key header.
  • Python + JavaScript SDKs on PyPI / npm as mintoken.
  • Mintoken CLI — run mintoken against any OpenAI-compatible endpoint from the terminal, including stdin → stdout streaming compression.
  • Browser extension — token usage badge for chatgpt.com, claude.ai, and gemini.google.com.
  • More tools beyond compression — semantic indexing, response caching, and smart model routing under the same plan.