Everything in the kit. Plan by plan.
What’s free forever and what Indie unlocks — each feature stamped with its plan, and the not-yet-shipped ones marked honestly as in build.
FULL INVENTORYKnow the terrain before you sign up
The maintained field manual: who offers free inference, on what terms, with what strings attached.
Public provider directory
Every free and free-tier inference provider we track — 48 and counting — searchable and comparable, including honest tombstones for providers that died or pivoted to paid.
Provider profiles
Per-provider dossiers: models, base URLs, OpenAI-compatibility, rate limits in the provider's own units (never coerced), and data-policy facts — does it train on your prompts, does it log — with citations.
Detailed friction metadata
The signup-pain matrix: exact steps to a key, card / phone / org requirements, time-to-first-key estimates. Know the cost in minutes before you burn them — free for everyone.
Weekly provider updates
A weekly ops loop re-checks sources and re-probes providers, so limits that changed on Tuesday don't burn you on Friday. Material changes land in the changelog.
Daily verified updates + drift alerts
The same verification loop on a daily cadence: daily catalog publishes over an authed feed, plus drift alerts when a limit, policy, or model materially moves — instead of finding out next week.
Why you can trust the numbers
Not scraped, not vibes. Every fact carries provenance; “unknown” means unknown.
Live probes
Three-level checks against the real APIs — reachability, model listings, and an actual completion. Listings lie (retired models linger in /models); probes are the ground truth.
Provenance on every claim
Each limit, policy, and price cites its source with a trust level and a verification stamp: verified, partially verified, unverified, or stale. No stamp inflation.
Source drift watch
Registered docs pages and curated repos are diffed for material changes — new limits, dead models, price moves — while cosmetic churn is ignored.
Versioned snapshots
Monthly catalog snapshots preserve history. Free always includes the current month. Indie opens the full archive, so what a provider promised in March is still on record in November.
From zero to key without the scavenger hunt
The onboarding hours are the product. We already spent them.
Signup & key-obtain guides
Step-by-step guides for every living provider we track, all 40: the exact console paths, the gotchas, the verification hoops. The free set is a rule, not a favor. Every provider that trains on your prompts gets a full free guide, plus four privacy-clean picks (OVHcloud, AI Horde, Requesty, Hugging Face), so privacy is never paywalled. Indie unlocks the whole shelf. Every gated guide shows its real first section and friction facts. No blur.
Key rotation runbooks
Per-provider procedures for rotating, testing, and retiring keys without downtime, matched to how each provider actually handles key lifecycles. Every guide carries its runbook, so the eight free guides include theirs.
Editorial “best for” guidance
Which models to actually reach for per job: writing, code, vision, long context. Raw capability flags stay free in the catalog. The curated judgment layer is Indie.
Proven wiring, not blank pages
Copy-paste starting points that already survived the edge cases.
Starter recipes
Working patterns to get moving: a free chat cascade, a gateway drop-in. Real code against real free tiers.
Full recipes library
The complete set, including the no-card/no-training privacy stack and vision on free tiers — each one probe-verified against current provider behavior.
Framework kits
Drop-in integration kits for LangChain, the OpenAI SDK, and the Vercel AI SDK — free-tier cascades wired into the framework you already use.
The control plane on your machine
Ships with the product, runs locally, BYOK. Your keys never leave your box.
The modelhubby CLI
list, show, recommend, friction, probe, health, keys, diff, snapshot, validate, doctor, serve — the whole catalog and verification engine from your terminal.
Local OpenAI-compatible gateway
One endpoint in front of all your free tiers: error-class-aware fallback, per-provider circuit breakers, SSE streaming passthrough. Point your app at localhost and stop hardcoding providers.
Cascade builder & code export
Policy-filtered, health-weighted provider cascades — exported as ready-to-paste code for your stack.
Server-built prebuilds
Cascade configs and code templates generated server-side from the live catalog. The open CLI can always build these locally from free data — Indie buys the pre-baked, always-current versions.
Key health & rotation
A nine-state key machine with error-class transitions and cooldowns. Keys are stored as fingerprints only; the schema physically rejects secret-shaped strings.
Watchlists & alerts
Watch the providers you depend on. Free: one watchlist with a weekly digest. Indie: 15 watches with instant email alerts the moment a limit, policy, or model materially moves.
