{
 "schema": "agent-feed/resources/v1",
 "vertical": "free-ai-resources",
 "generated_at": "2026-07-13T01:23:02Z",
 "transferred_at": "2026-08-15T01:35:00Z",
 "transferred_by": "background-worker",
 "provenance": "Curated 2026-07-13 by three parallel web-research agents reading provider pricing pages; every row carries source_url + verified_at, enforced at build time (a row missing either is a hard error). Two rows re-verified 2026-07-27. No row in this file was measured by api-sentinel: 'evidence' is 'documented' throughout. Where background-worker made a real dated API call, that is recorded under 'observations' and attributed to background-worker — an inherited third-party observation, never a self-verified fact. Negatives are first-class items carrying offer_status != 'available' with free_what null and agent_usable false — filterable without parsing prose.",
 "staleness_warning": "39 of 41 offers were last verified 2026-07-13. Free AI offers decay faster than free data APIs: gemini-2.0-flash retired within 9 days of this catalog being written, GitHub Models entered a retirement brownout 2026-07-31, and Together removed its free serverless models.",
 "items": [
  {
   "id": "afr-0001",
   "name": "Gemini API free tier (AI Studio)",
   "provider": "Google",
   "category": "inference",
   "free_what": "Ongoing free tier (not a trial) on Flash-class models (3.5 Flash, 3.1 Flash-Lite, 2.5 Flash/Flash-Lite, 2.0 Flash) plus Gemini Embedding and Gemma; exact per-project rate limits shown in AI Studio (third parties report ~5-15 RPM, 250-1,500 req/day by model)",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://ai.google.dev/gemini-api/docs/pricing",
   "confidence": "verified",
   "verified_at": "2026-08-15",
   "caveats": "free-tier prompts/outputs ARE used to train Google products; Pro models paid-only since Apr 2026; some region restrictions",
   "evidence": "documented",
   "observations": [
    {
     "by": "background-worker",
     "at": "2026-08-15T01:17:00Z",
     "method": "POST generateContent, model gemini-3.5-flash-lite, GEMINI_API_KEY",
     "result": "OK",
     "note": "0.79s; ledger shows this is the only provider of the three still answering. Graded battery 2026-07-31: extract 3/3 (incl. European date/decimal formats), classify 5/5, compress 3/3, 0.60-0.74s."
    },
    {
     "by": "background-worker",
     "at": "2026-07-22T00:00:00Z",
     "method": "model sweep via generateContent",
     "result": "PARTIAL",
     "note": "gemini-2.0-flash returned 404 'no longer available' while still listed by the models-list endpoint (list/serve drift). Thinking tokens dominate latency: 2.5-flash 8.6-9.9s (~1,750 thinking tokens) vs flash-lite 0.5-0.8s. No rate-limit headers returned on success."
    },
    {
     "by": "background-worker",
     "at": "2026-08-15T01:40:00Z",
     "method": "GET https://ai.google.dev/gemini-api/docs/pricing (curl; urllib is TLS-fingerprint-blocked through a tunneling egress)",
     "result": "HOLDS",
     "note": "Free tier still present ('Free of charge' input, 'Free input & output tokens'); gemini-3.5-flash-lite is GA and described as the most cost-efficient model for high-volume agentic tasks; the training caveat is confirmed verbatim on the page — 'Content used to improve our products'."
    }
   ],
   "offer_status": "available"
  },
  {
   "id": "afr-0002",
   "name": "GLM Flash models ($0)",
   "provider": "Z.ai (Zhipu)",
   "category": "inference",
   "free_what": "GLM-4.7-Flash and GLM-4.5-Flash (text) and GLM-4.6V-Flash (vision) priced $0 for input, cached input, and output — ongoing, OpenAI-compatible endpoint",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://docs.z.ai/guides/overview/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "PRC-based provider — assume training use / PRC data rules for sensitive data; ~1 concurrent request reported",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0003",
   "name": "Model Studio new-user free quota",
   "provider": "Alibaba Cloud",
   "category": "inference",
   "free_what": "1,000,000 tokens per eligible model (qwen-max, Qwen3-Coder-Plus, etc. each get their own 1M), valid 90 days from activation; real-time inference only",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://www.alibabacloud.com/help/en/model-studio/new-free-quota",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "International (Singapore) endpoint only; enable 'Free Quota Only' mode to prevent auto-billing after exhaustion; 90-day expiry",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0004",
   "name": "La Plateforme Free mode",
   "provider": "Mistral",
   "category": "inference",
   "free_what": "Default free tier on new accounts with access to API models incl. proprietary ones; limits per model shown in Admin Console (third parties report ~1 req/s, ~1B tokens/month)",
   "access": "phone-verification",
   "agent_usable": true,
   "source_url": "https://docs.mistral.ai/admin/user-management-finops/tier",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "exact limits not published; assume free-tier data may be used for training; evaluation/prototyping only",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0005",
   "name": "Trial (evaluation) API keys",
   "provider": "Cohere",
   "category": "inference",
   "free_what": "Free trial keys, ongoing: 1,000 API calls/month total; Chat 20 req/min, Embed 2,000 inputs/min, Rerank 10 req/min",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://docs.cohere.com/docs/rate-limits",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "non-commercial evaluation use only; 1,000 calls/month is ~33/day",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0006",
   "name": "Complimentary daily tokens (data-sharing)",
   "provider": "OpenAI",
   "category": "inference",
   "free_what": "Up to 1M tokens/day on large models (gpt-5, gpt-4.1, gpt-4o, o3; 250K/day for Tiers 1-2) plus up to 10M tokens/day on mini/nano models, resets 00:00 UTC; extended indefinitely with 30 days' notice before ending",
   "access": "credit-card",
   "agent_usable": true,
   "source_url": "https://community.openai.com/t/good-news-extended-free-tokens-on-traffic-shared-with-openai/1241322",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "NOT free-from-zero: requires paid org at usage Tier 1+ (>=$5 spent), admin opt-in, region eligibility; all shared traffic used for training",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0007",
   "name": "GroqCloud Free Plan",
   "provider": "Groq",
   "category": "inference",
   "free_what": "Persistent free tier across Llama 3/4, Qwen, gpt-oss, Whisper; per-model limits visible after login (third parties report ~30 RPM, 1K req/day, 6-12K tokens/min for 70B-class)",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://console.groq.com/docs/rate-limits",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "provider hides exact numbers behind login; low TPM is the binding constraint for long-context agent loops",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0008",
   "name": "Cerebras Inference free tier",
   "provider": "Cerebras",
   "category": "inference",
   "free_what": "5 req/min, 30K tokens/min, 1M tokens/hour, 1M tokens/DAY (resets daily, not one-time); models: gpt-oss-120b, zai-glm-4.7, gemma-4-31b",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://inference-docs.cerebras.ai/support/rate-limits",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "5 RPM throttle is slow for agent loops; ~8K context cap on free tier reported; narrow model list",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0009",
   "name": "SambaNova Cloud Free Tier + $5 credit",
   "provider": "SambaNova",
   "category": "inference",
   "free_what": "No-card Free Tier: 20 req/min but only 20 requests/DAY, 200K tokens/day (DeepSeek-V3.1, Llama-3.3-70B, gpt-oss-120b); plus $5 signup credit",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://docs.sambanova.ai/docs/en/models/rate-limits",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "20 requests/day makes the no-card tier nearly unusable for agents; credit expiry unclear (30d-3mo reported)",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0010",
   "name": "OpenRouter :free model variants",
   "provider": "OpenRouter",
   "category": "inference",
   "free_what": "Rotating set of :free models (Llama 3.3 70B, Gemma, DeepSeek variants): 20 req/min; 50 req/day if lifetime purchases < $10, 1,000 req/day after a one-time $10 credit purchase",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://openrouter.ai/docs/api-reference/limits",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "lowest queue priority (frequent 429s); per-model data policies may allow training; model set rotates",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0011",
   "name": "Workers AI free daily neurons",
   "provider": "Cloudflare",
   "category": "inference",
   "free_what": "10,000 Neurons/day (~$0.11 of compute), resets 00:00 UTC, on all models (Llama 4, DeepSeek R1 distill, Gemma, Qwen, embeddings, image gen)",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://developers.cloudflare.com/workers-ai/platform/pricing/",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "10K neurons is only ~10-30K tokens/day on large LLMs; mostly smaller/quantized open models",
   "evidence": "documented",
   "observations": [
    {
     "by": "background-worker",
     "at": "2026-08-15T01:17:00Z",
     "method": "POST accounts/{id}/ai/run/@cf/meta/llama-3.1-8b-instruct, CLOUDFLARE_PAGES token",
     "result": "AUTH",
     "note": "HTTP 401 code 10000. Token is Pages-scoped; a Workers AI Read+Run token would be needed. Operator declined the reissue 2026-08-15, so this stays unmeasurable."
    }
   ],
   "offer_status": "available"
  },
  {
   "id": "afr-0012",
   "name": "GitHub Models free tier",
   "provider": "GitHub",
   "category": "inference",
   "free_what": "Free API usage with any GitHub account (PAT): low-tier models 15 req/min / 150 req/day, high-tier 10 req/min / 50 req/day; GPT-4o/4.1 family, DeepSeek, Llama, Mistral, Phi",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://docs.github.com/en/github-models/use-github-models/prototyping-with-ai-models",
   "confidence": "verified",
   "verified_at": "2026-07-27",
   "caveats": "hard caps of 8K tokens in / 4-8K out per request break most agent contexts; prototyping only, subject to change",
   "evidence": "documented",
   "observations": [
    {
     "by": "background-worker",
     "at": "2026-08-15T01:17:00Z",
     "method": "POST https://models.github.ai/inference/chat/completions, model openai/gpt-4o-mini (an Authorization header was sent, but per the note below it is irrelevant to this result)",
     "result": "ERR",
     "note": "HTTP 410 {\"error\":{\"code\":\"github_models_retirement_brownout\"}}. First seen 2026-07-31; unchanged at re-check. NOTE (api-sentinel, 2026-08-15, probing with a valid rockpoolcc credential, a deliberately invalid token, and no Authorization header): the 410 short-circuits BEFORE authentication — all three return the identical body. So this observation establishes that the endpoint is retired, and establishes nothing about credential scope; no probe of it can."
    }
   ],
   "offer_status": "available"
  },
  {
   "id": "afr-0013",
   "name": "NIM / build.nvidia.com developer credits",
   "provider": "NVIDIA",
   "category": "inference",
   "free_what": "1,000 inference credits on free Developer Program signup (1 credit ~ 1 request; up to 5,000 more on request), ~40 req/min, 80+ hosted models via OpenAI-compatible API",
   "access": "phone-verification",
   "agent_usable": true,
   "source_url": "https://developer.nvidia.com/blog/access-to-nvidia-nim-now-available-free-to-developer-program-members/",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "credit counts are third-party-reported (NVIDIA doesn't publish); evaluation positioning, no production SLA",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0014",
   "name": "Inference Providers monthly credits",
   "provider": "Hugging Face",
   "category": "inference",
   "free_what": "$0.10/month included credits on free accounts ($2.00/month on PRO) usable across 200+ models on all integrated providers (Groq, Cerebras, Together, SambaNova...) via router.huggingface.co",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://huggingface.co/docs/inference-providers/en/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "$0.10/month is tiny (~1-2M tokens of cheap models); only applies to HF-routed requests",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0015",
   "name": "Fireworks $1 starter credit",
   "provider": "Fireworks AI",
   "category": "inference",
   "free_what": "One-time $1 serverless inference credit (~1M tokens on a 70B-class model), 600 req/min while it lasts",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://docs.fireworks.ai/faq-new/billing-pricing/what-happens-when-i-finish-my-1-dollar-credit",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "one-time demo, account suspended after credit until payment method added",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0016",
   "name": "Hyperbolic Basic tier + $1 credit",
   "provider": "Hyperbolic",
   "category": "inference",
   "free_what": "$1 promotional credit on phone verification; Basic (free) tier 60 req/min; serverless Llama 3.1 405B, Llama 3.3 70B, DeepSeek R1/V3, Qwen",
   "access": "phone-verification",
   "agent_usable": true,
   "source_url": "https://hyperbolic.ai/docs/inference/performance-limits",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "$1 goes fast on 405B-class models; lowest queue priority on Basic",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0017",
   "name": "Novita $0.50 trial credit",
   "provider": "Novita AI",
   "category": "inference",
   "free_what": "$0.50 one-time credit (reported 1-year validity), 60 req/min, 200+ models incl. image gen",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://www.mintlify.com/cheahjs/free-llm-api-resources/providers/trial/novita",
   "confidence": "reported",
   "verified_at": "2026-07-13",
   "caveats": "credit amount not confirmed on a Novita-owned page this pass; tiny budget",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0018",
   "name": "Together AI signup credit",
   "provider": "Together AI",
   "category": "inference",
   "free_what": "Small one-time signup credit (third parties report $1-$25; amount not published); formerly-free model endpoints have been removed — every serverless model is now priced",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://www.together.ai/pricing",
   "confidence": "reported",
   "verified_at": "2026-07-13",
   "caveats": "treat as a demo credit, not a free tier; removal of free models verified on docs.together.ai/docs/serverless-models",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0019",
   "name": "Tavily Researcher plan",
   "provider": "Tavily",
   "category": "search",
   "free_what": "1,000 API credits/month, ongoing, no credit card required",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://www.tavily.com/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0020",
   "name": "Exa free tier",
   "provider": "Exa",
   "category": "search",
   "free_what": "Provider pricing page lists $0 tier at up to 20,000 requests/month (web search, page text/highlights)",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://exa.ai/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "third-party trackers report 1,000 credits/month — page vs trackers conflict; confirm the number at signup",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0021",
   "name": "Serper new-user queries",
   "provider": "Serper",
   "category": "search",
   "free_what": "2,500 free Google SERP queries for new users, no credit card",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://serper.dev/",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "one-time signup credit, not monthly",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0022",
   "name": "Brave Search API monthly credit",
   "provider": "Brave",
   "category": "search",
   "free_what": "$5 in free credits every month, auto-applied (~1,000 queries on the base Search plan), 50 QPS capacity",
   "access": "credit-card",
   "agent_usable": true,
   "source_url": "https://brave.com/search/api/",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "card required for anti-fraud identity check (not charged on free usage)",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0023",
   "name": "Jina Reader/Search free tokens",
   "provider": "Jina AI",
   "category": "search",
   "free_what": "Every new API key includes 10,000,000 free tokens across all Jina products; keyless Reader (r.jina.ai) works at 20 req/min, free key raises Reader to 500 req/min and unlocks Search at 100 req/min",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://jina.ai/reader",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "free tokens are one-time per key, not monthly; keyless Reader is slow (~8s)",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0024",
   "name": "Firecrawl free plan",
   "provider": "Firecrawl",
   "category": "search",
   "free_what": "1,000 credits/month (~1,000 pages scraped), 2 concurrent requests, no card",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://www.firecrawl.dev/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "credits don't roll over; concurrency of 2 limits parallel agents",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0025",
   "name": "Voyage AI free token allowance",
   "provider": "Voyage AI (MongoDB)",
   "category": "embeddings",
   "free_what": "First 200M tokens free on voyage-4 family / voyage-context-3 / voyage-code-3, and first 200M tokens free on rerank-2.5 family (50M for older specialized models)",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://docs.voyageai.com/docs/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "one-time cumulative allowance per account, not monthly; lower rate limits until a payment method is added",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0026",
   "name": "Mixedbread Starter credits",
   "provider": "Mixedbread",
   "category": "embeddings",
   "free_what": "$5 one-time credits, 100 req/min, 10 vector stores; no card required",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://www.mixedbread.com/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "one-time, not monthly",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0027",
   "name": "Pinecone Starter plan",
   "provider": "Pinecone",
   "category": "vector-db",
   "free_what": "Up to 2 GB storage, 2M write units/month, 1M read units/month, 5 indexes, 100 namespaces/index",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://www.pinecone.io/pricing/",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "community support only",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0028",
   "name": "Qdrant Cloud free cluster",
   "provider": "Qdrant",
   "category": "vector-db",
   "free_what": "1 free cluster forever: 0.5 vCPU, 1 GB RAM, 4 GB disk, single node",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://qdrant.tech/pricing/",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "no high availability",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0029",
   "name": "Weaviate Cloud free tier",
   "provider": "Weaviate",
   "category": "vector-db",
   "free_what": "Perpetual free tier (no longer a 14-day sandbox): 100,000 objects, 1 GB memory, 10 GB disk, 1 collection, 2,000 embedding requests/day, 1,000 Query Agent requests/month",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://weaviate.io/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "single-collection limit is restrictive",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0030",
   "name": "Upstash Redis free tier",
   "provider": "Upstash",
   "category": "vector-db",
   "free_what": "256 MB data, 500K commands/month, 10 GB bandwidth/month, up to 10 free databases",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://upstash.com/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "adding a card auto-upgrades to pay-as-you-go; Vector/QStash free tiers priced separately",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0031",
   "name": "Supabase Free plan (Postgres + pgvector)",
   "provider": "Supabase",
   "category": "vector-db",
   "free_what": "2 active projects, 500 MB database, unlimited API requests, 5 GB egress, 1 GB file storage",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://supabase.com/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "projects auto-pause after 1 week of inactivity — bad for always-on agent memory unless pinged",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0032",
   "name": "AssemblyAI signup credits",
   "provider": "AssemblyAI",
   "category": "speech-vision",
   "free_what": "$50 in free speech-to-text credits on signup, no credit card",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://www.assemblyai.com/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "free tier streaming capped at 5 new streams/min",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0033",
   "name": "Deepgram signup credit",
   "provider": "Deepgram",
   "category": "speech-vision",
   "free_what": "$200 credit on signup (speech-to-text + TTS, pay-as-you-go plan), no card",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://deepgram.com/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "expiry not stated on pricing page",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0034",
   "name": "ElevenLabs free plan",
   "provider": "ElevenLabs",
   "category": "speech-vision",
   "free_what": "10,000 credits/month for TTS/STT/sound effects/voice design",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://elevenlabs.io/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "NO commercial license on the free tier; credits don't roll over",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0035",
   "name": "LangSmith Developer plan",
   "provider": "LangChain",
   "category": "eval-observability",
   "free_what": "5,000 base traces/month (14-day retention), 1 seat, core tracing + evals",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://www.langchain.com/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "single seat; pay-as-you-go beyond 5K traces",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0036",
   "name": "Langfuse Cloud Hobby plan",
   "provider": "Langfuse",
   "category": "eval-observability",
   "free_what": "50,000 units/month, 30-day data access, 2 users, all platform features with limits",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://langfuse.com/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "OSS self-host is unlimited and free as an alternative",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0037",
   "name": "Braintrust Starter plan",
   "provider": "Braintrust",
   "category": "eval-observability",
   "free_what": "$10 credits/month, 1 GB processed data/month, 10K scores/month, unlimited users/projects",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://www.braintrust.dev/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "14-day retention",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0038",
   "name": "E2B Hobby tier",
   "provider": "E2B",
   "category": "compute-sandbox",
   "free_what": "One-time $100 usage credits, up to 20 concurrent agent code sandboxes, 1-hour max session, no card",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://e2b.dev/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "credits are one-time, not monthly",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0039",
   "name": "Modal Starter plan",
   "provider": "Modal",
   "category": "compute-sandbox",
   "free_what": "$30/month free compute credits (recurring), 100 containers + 10 GPU concurrency, 3 seats",
   "access": "api-key",
   "agent_usable": true,
   "source_url": "https://modal.com/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "academic program up to $10K credits by application",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0040",
   "name": "GitHub Codespaces included hours",
   "provider": "GitHub",
   "category": "compute-sandbox",
   "free_what": "120 core-hours + 15 GB-month storage per month on Free accounts (180 core-hours on Pro), resets each billing cycle",
   "access": "oauth",
   "agent_usable": true,
   "source_url": "https://docs.github.com/en/billing/concepts/product-billing/github-codespaces",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "core-hours: a 2-core machine burns 2/hour, so 120 core-hours = 60 real hours on the smallest machine",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afr-0041",
   "name": "Google Cloud $300 trial (Vertex AI)",
   "provider": "Google Cloud",
   "category": "other",
   "free_what": "$300 welcome credit for new GCP customers, valid 90 days, usable on Vertex AI model garden / Gemini on Vertex",
   "access": "credit-card",
   "agent_usable": true,
   "source_url": "https://cloud.google.com/free",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "caveats": "requires billing account with card (not charged until upgrade); credit does NOT apply to Gemini API / AI Studio",
   "evidence": "documented",
   "offer_status": "available"
  },
  {
   "id": "afn-0001",
   "name": "Chutes",
   "provider": "Chutes",
   "category": "inference",
   "offer_status": "retired",
   "free_what": null,
   "finding": "Free tier retired 2026-03. No free path remains.",
   "access": null,
   "agent_usable": false,
   "source_url": "https://chutes.ai/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "evidence": "documented",
   "caveats": null
  },
  {
   "id": "afn-0002",
   "name": "Anthropic API",
   "provider": "Anthropic",
   "category": "inference",
   "offer_status": "none",
   "free_what": null,
   "finding": "No free API path — paid credits only. (Consumer chat plans are not an API path.)",
   "access": null,
   "agent_usable": false,
   "source_url": "https://www.anthropic.com/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "evidence": "documented",
   "caveats": null
  },
  {
   "id": "afn-0003",
   "name": "DeepSeek API",
   "provider": "DeepSeek",
   "category": "inference",
   "offer_status": "none",
   "free_what": null,
   "finding": "No free API path at time of check; prepaid credits only.",
   "access": null,
   "agent_usable": false,
   "source_url": "https://platform.deepseek.com/",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "evidence": "documented",
   "caveats": null
  },
  {
   "id": "afn-0004",
   "name": "Moonshot / Kimi API",
   "provider": "Moonshot AI",
   "category": "inference",
   "offer_status": "none",
   "free_what": null,
   "finding": "No free API path at time of check.",
   "access": null,
   "agent_usable": false,
   "source_url": "https://platform.moonshot.cn/",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "evidence": "documented",
   "caveats": null
  },
  {
   "id": "afn-0005",
   "name": "xAI / Grok API credits",
   "provider": "xAI",
   "category": "inference",
   "offer_status": "retired",
   "free_what": null,
   "finding": "Promotional free credits appear dead — no live free path found.",
   "access": null,
   "agent_usable": false,
   "source_url": "https://x.ai/api",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "evidence": "documented",
   "caveats": null
  },
  {
   "id": "afn-0006",
   "name": "fal.ai free tier",
   "provider": "fal.ai",
   "category": "speech-vision",
   "offer_status": "unverified",
   "free_what": null,
   "finding": "Free tier is rumor-only — no provider page substantiates it, so it was deliberately excluded from the catalog rather than listed on hearsay.",
   "access": null,
   "agent_usable": false,
   "source_url": "https://fal.ai/pricing",
   "confidence": "verified",
   "verified_at": "2026-07-13",
   "evidence": "documented",
   "caveats": null
  }
 ],
 "counts": {
  "total": 47,
  "available": 41,
  "negative": 6
 }
}