OpenAI's Decisions API on its own wire: point an OpenAI SDK's base URL here and client.decisions.create works unchanged, paid per call. Send input (text, or user messages with input_text parts) and questions; get back answers your code can branch on. predicate returns the probability a condition holds; choice returns one of your values with probabilities and confidence; score returns a probability-weighted position on your ordered levels. Up to 8 questions per call.

Hand over a domain and get one graded security & email-deliverability audit: SPF, DMARC, DKIM and MX (why your mail lands in spam), the web security headers, and the TLS certificate - every finding from a live probe, with an overall letter grade, a downloadable checks appendix, and a prioritized, specific list of fixes. USDC (x402/MPP) or card (Stripe). Not cached.

The deeper tier: everything in the standard audit plus the attack surface from Certificate Transparency logs (subdomains), the detected tech stack, and domain registration, in a longer graded report with a fuller remediation plan. USDC or card (Stripe). Not cached.

Hand over a ticker and get one due-diligence dossier back. Pulls the company's SEC EDGAR filings (10-K / 10-Q / 8-K) and Form 4 insider trades, a live quote, and grounded web research on recent developments and risks, then synthesizes a ~2,400-word cited dossier - business, financial posture, insider activity, red flags, and a balanced diligence read. Grounded in real SEC data, not a generic web guess. One payment, one report. USDC (x402/MPP) or card (Stripe). Not cached.

The exhaustive due-diligence tier: more filings, a full year of Form 4 insider activity, wider web research (up to 8 angles), the widest source set, and a ~2,800-word cited dossier with a full source table. For a decision that warrants real diligence. USDC or card (Stripe). Not cached.

OpenAI-compatible text embeddings (text-embedding-3-small by default; 3-large and ada-002 supported), up to 64 inputs or 16k chars per call, returned in the standard OpenAI shape. Identical inputs repeated within 10 minutes are served free from cache. Use it for semantic search, clustering and retrieval from any OpenAI SDK by changing base_url.

Name a US ticker and get one cited report on what the company just filed with the SEC: every filing in EDGAR's index for the window (8-K, 10-Q, 10-K, S-1, 424B4, DEF 14A and the rest, minus any form you exclude), the primary documents of the most consequential ones read in full and explained in plain language, and what changed versus the prior period where the filing itself says so, with a downloadable filings appendix. Not investment advice. Not cached.

Hand over an institutional manager (a fund name like "Berkshire Hathaway", a ticker, or a SEC CIK) and get one cited report on what they hold and what they BOUGHT, ADDED, TRIMMED, and EXITED last quarter - diffed from their two most recent SEC 13F-HR filings, with a downloadable holdings + changes appendix. Data a chatbot can't reach (parsed live off SEC EDGAR). One payment, one report. USDC (x402/MPP) or card (Stripe). Not cached.

The deep tier: the full holdings table, wider quarter-over-quarter change analysis, more grounded web research, and a longer cited report on the manager's positioning. USDC or card (Stripe). Not cached.

Google's native generateContent wire on the base tier. Send Gemini's own request shape (contents, systemInstruction, generationConfig, function declarations) and get Gemini's own response shape (candidates, usageMetadata) back. Point a Google GenAI SDK at this gateway and pay per request with USDC, no account. Also answers the Google-shaped path /v1beta/models/<model>:generateContent.

OpenAI-compatible chat completions GROUNDED in a live web search on every call: the gateway runs an Exa search (up to 5 results) for the prompt, hands the results to the model, and returns the answer with url_citation annotations - $0.03 per call in USDC, no API key, no signup. Model is chosen server-side like the auto tier (omit "model" or send "auto"; explicit ranked models accepted); the response adds agent402_router {category, quality, served}.

Generate an image from a text prompt at a flat price per picture, no token math and no subscription: the OpenAI images request shape in, base64 PNG or JPEG out, with an automatic fallback model so a provider outage does not become your error. Use it when an agent needs a picture cheaply and predictably.

OpenAI-compatible image generation over x402 - point any OpenAI SDK's images.generate() at base_url https://agent402.tools/v1 and pay $0.08 per image in USDC, no API key, no signup. Served by FLUX.2 Pro with GPT-5 Image Mini as the failover; prompt in (up to 4k chars), one 1024x1024 PNG out as inline base64 (response_format b64_json). One image per call (n locked to 1). Optional zdr:true routes only to zero-data-retention providers.

Higher-fidelity text-to-image on the OpenAI images wire: a prompt in, one 1024x1024 image out as inline base64, served by a pro-grade diffusion model in about ten seconds. Use it when an agent needs a finished picture for a page, a post or a product and quality matters more than the cheapest draft.

Name a US ticker and get one cited insider-flow report: every Form 4 against the company in the window with the actual transactions parsed from the filings - open-market buys and sales (who, when, how much, at what price) separated from awards, option exercises and tax withholding - aggregated per insider and as a net flow, with a grounded signal read and a downloadable transactions appendix. SEC EDGAR data, not investment advice. USDC (x402/MPP) or card (Stripe). Not cached.

A deterministic digest of the IPO pipeline for a window: every 424B4 final prospectus and every S-1/F-1 registration from SEC EDGAR, each filer classified from its own submissions index as a first-time registrant (an IPO), an already-public issuer (follow-on or resale, not an IPO) or a SPAC, optionally filtered by a keyword on the filer's name, with tickers/exchanges where listed and a downloadable filings table. No LLM, no guessing - filing facts only. Default last 7 days.

Ask a typed question about any content and get an answer your code can branch on, not prose to parse. Send state (the content, a string or object) and questions (your own ids mapped to question objects); get back answers keyed by those ids. Three question types: choice (pick one of your named options: returns choice, probabilities per option and confidence), score (a position on levels you describe, lowest first: returns score as a fractional level index, probabilities per level index.

Hand over a topic and get a publish-ready LinkedIn article back: grounded web research with cited sources, three headline options, a hook-first body under subheads with facts linked to their sources, key takeaways, a companion post with hashtags, and generated images delivered at LinkedIn's own sizes (article cover 1920x1080, link-share 1200x627, feed square and portrait, in-article 16:9) as inline JPEG under 3 MB. Optional byline, angle, audience, tone and CTA.

Anthropic Messages API over x402 - point the Anthropic SDK (or Claude Code / the Agent SDK) at base_url https://agent402.tools/v1 and pay $0.02 per call in USDC, no API key, no signup. Same models, caps and price as this tier's /chat/completions route; any model here is served through the Messages wire (Claude natively, others translated). Depth: `effort` (low..max) on Claude 4.7+, thinking.budget_tokens on older Claude. Up to 32,000 input chars and 2048 output tokens; streaming supported.

OpenAI-compatible chat completions priced from the request itself: the 402 quotes this exact body, from $0.001, and a buyer on the upto scheme settles the tokens actually used under that quote. Use it when an agent sends large or variable prompts and wants to pay for what it sends, not a flat tier.

Google's generateContent wire priced from the request itself: the 402 quotes this exact body, from $0.001, with upto buyers settling actual usage. Use it when an agent built on the Gemini SDK needs a supported model paid per call without an API key.

The Anthropic Messages wire priced from the request itself: the 402 quotes this exact body, from $0.001, with upto buyers settling actual usage under the quote. Use it when an agent built on the Anthropic SDK, or Claude Code, needs any supported model paid per call with no account.

The OpenAI Responses wire priced from the request itself: each 402 quotes this exact body, from $0.001, and upto buyers settle actual usage under it. Use it when an agent built on the Responses API or the OpenAI Agents SDK wants per-call payment sized to what it sends.

OpenAI-compatible chat completions, nano tier: gpt-6-luna (the default), gpt-5.6-luna, gpt-5-nano, gemini flash-lite, small llama/ministral/qwen, deepseek-chat - $0.003 per call in USDC over x402, priced for high-frequency agent loops. Same wire format as /v1/chat/completions with loop-sized caps (12k chars in, 768 tokens out). Streaming supported (stream: true). No API key, no signup.

Google's native generateContent wire on the nano tier. Send Gemini's own request shape (contents, systemInstruction, generationConfig, function declarations) and get Gemini's own response shape (candidates, usageMetadata) back. Point a Google GenAI SDK at this gateway and pay per request with USDC, no account. Also answers the Google-shaped path /v1beta/models/<model>:generateContent.

Anthropic Messages API over x402 - point the Anthropic SDK (or Claude Code / the Agent SDK) at base_url https://agent402.tools/v1/nano and pay $0.003 per call in USDC, no API key, no signup. Same models, caps and price as this tier's /chat/completions route; any model here is served through the Messages wire (Claude natively, others translated). Depth: `effort` (low..max) on Claude 4.7+, thinking.budget_tokens on older Claude. Up to 12,000 input chars and 768 output tokens; streaming supported.

OpenAI Responses API over x402 - point the OpenAI SDK's responses.create() (or the OpenAI Agents SDK) at base_url https://agent402.tools/v1/nano and pay $0.003 per call in USDC, no API key, no signup. Same models, caps and price as this tier's /chat/completions route; any model here is served through the Responses wire.

OpenAI-compatible chat completions, premium tier: gpt-5, gpt-6 astra, o3 and o4-mini, claude opus, claude fable 5.1 - paid per call in USDC over x402. Same wire format as /v1/chat/completions with the largest caps (85k chars in, 8192 tokens out).

Google's native generateContent wire on the premium tier. Send Gemini's own request shape (contents, systemInstruction, generationConfig, function declarations) and get Gemini's own response shape (candidates, usageMetadata) back. Point a Google GenAI SDK at this gateway and pay per request with USDC, no account. Also answers the Google-shaped path /v1beta/models/<model>:generateContent.

Anthropic Messages API over x402 - point the Anthropic SDK (or Claude Code / the Agent SDK) at base_url https://agent402.tools/v1/premium and pay $0.50 per call in USDC, no API key, no signup. Same models, caps and price as this tier's /chat/completions route; any model here is served through the Messages wire (Claude natively, others translated). Depth: `effort` (low..max) on Claude 4.7+, thinking.budget_tokens on older Claude.

OpenAI Responses API over x402 - point the OpenAI SDK's responses.create() (or the OpenAI Agents SDK) at base_url https://agent402.tools/v1/premium and pay $0.50 per call in USDC, no API key, no signup. Same models, caps and price as this tier's /chat/completions route; any model here is served through the Responses wire.

OpenAI's transcription wire on the ten-minute tier: POST multipart/form-data with a `file` part. Same model as /v1/audio/transcriptions with a longer cap, for a recording a four-minute route refuses.

OpenAI-compatible chat completions, pro tier: gpt-4o, gpt-4.1, claude sonnet, gemini pro, grok - paid per call in USDC over x402. Same wire format as /v1/chat/completions with higher input/output caps (48k chars in, 4096 tokens out).

Google's native generateContent wire on the pro tier. Send Gemini's own request shape (contents, systemInstruction, generationConfig, function declarations) and get Gemini's own response shape (candidates, usageMetadata) back. Point a Google GenAI SDK at this gateway and pay per request with USDC, no account. Also answers the Google-shaped path /v1beta/models/<model>:generateContent.

Anthropic Messages API over x402 - point the Anthropic SDK (or Claude Code / the Agent SDK) at base_url https://agent402.tools/v1/pro and pay $0.10 per call in USDC, no API key, no signup. Same models, caps and price as this tier's /chat/completions route; any model here is served through the Messages wire (Claude natively, others translated). Up to 48,000 input chars and 4096 output tokens; streaming supported.

OpenAI Responses API over x402 - point the OpenAI SDK's responses.create() (or the OpenAI Agents SDK) at base_url https://agent402.tools/v1/pro and pay $0.10 per call in USDC, no API key, no signup. Same models, caps and price as this tier's /chat/completions route; any model here is served through the Responses wire.

Name a drug, food, ingredient, brand or medical device and get one cited FDA recall report: every matching enforcement record across the drug, food and device feeds (firm, class I/II/III, status, reason, distribution, date), organized and explained, with a downloadable records appendix. Live openFDA data, not medical advice. USDC (x402/MPP) or card (Stripe). Not cached.

Rerank documents against a query over x402 - the Cohere /rerank wire ({query, documents[], top_n} -> results with relevance_score), served by cohere/rerank-v3.5, $0.002 per call in USDC, no API key, no signup. Up to 50 documents (1,600 chars each, 40k total) and a 500-char query per call.

Hand over a whole research question and get one cited report back. The gateway plans sub-questions, runs multiple live web searches, reranks the sources by relevance, and synthesizes a ~1,500-word report with inline [n] citations and a source list - one payment, one outcome. Priced per report, not per call. Payable in USDC (x402/MPP) or by card (Stripe, >= $0.50 minimum - this clears it). Not cached (the web moves).

Name a market, category or company and get one cited MARKET / COMPETITOR BRIEF: how the market is defined and sized, the key players and what each offers, pricing, recent moves, how they differ, risks and a bottom line - every claim from live web research with citations, nothing from memory. The research pipeline with a competitive-intelligence plan. USDC (x402/MPP) or card (Stripe). Not cached.

The exhaustive tier: up to a dozen sub-questions and grounded searches, the widest reranked source set, premium synthesis into a ~2,800-word cited report with a full source table. Our most thorough single-call research report. USDC or card (Stripe). Not cached.

The deeper research tier: more sub-questions, more grounded searches, wider reranked source set, and a premium (Claude Opus class) synthesis into a ~2,200-word cited report with structured findings. For questions worth a real dossier. USDC or card (Stripe). Not cached.

OpenAI Responses API over x402 - point the OpenAI SDK's responses.create() (or the OpenAI Agents SDK) at base_url https://agent402.tools/v1 and pay $0.02 per call in USDC, no API key, no signup. Same models, caps and price as this tier's /chat/completions route; any model here is served through the Responses wire.

One ticker, one purchase, the full picture. Runs the company due-diligence dossier (SEC filings, XBRL financials, grounded web research), the Form 4 insider-flow report with the actual transactions parsed, and a 13F institutional-holder sample read live off SEC EDGAR full-text search, then stitches them into ONE cited report: company, filings, insider flow, institutional holders and what to check next, with a merged source list and downloadable data appendices.

Hand over a Solana SPL mint address and get one cited due-diligence brief: what the token is, whether the mint and freeze authorities are still live, how deep and how locked the liquidity is (pools, lockers, unlock dates, DEX pairs, 24h volume and buy/sell balance), who holds it (top holders with pool and locker accounts told apart from wallets, concentration with and without pools), and every named risk flag the public feeds carry, plus what this data cannot see.

Hand over a token contract address (and chain) and get one evidence-based on-chain risk report: whether the source is verified, holder concentration (top-1 / top-10 share of supply, distinguishing pools/contracts from wallets), and supply + market context, with a top-holders appendix. Evidence, never a 'safe' or 'scam' verdict - on-chain checks can't see off-chain rugs, so a clean report is not an endorsement. USDC (x402/MPP). Not cached.

The deeper tier: everything in the standard report plus a deterministic static-pattern scan of the verified source (tx.origin auth, delegatecall, selfdestruct, unchecked calls, reentrancy surface, etc. - heuristic triage, not a formal audit), more holders, and a web reputation check. Still evidence, never a verdict. USDC (x402/MPP). Not cached.

Generate a short video clip from a text prompt at a flat price per clip: a four-second 720p clip returned inline as base64 MP4, with the duration and resolution locked so the price never moves. Use it when an agent needs motion rather than a still, and needs to know the cost before it calls.