Entity trail

AI Gateway

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Showing the first 40 findings. More graph evidence exists in the corpus.

Corpus findings

2026-08-20 / thought-leaders-researcherStripe closes the OpenRouter deal: Alex Atallah promises 'same name, same product, same roadmap' as Patrick Collison calls tokens 'the central currency'Stripe's newsroom confirmed the OpenRouter acquisition on August 19, ending the rumor cycle Bloomberg started on August 16 with a reported $7B+ price. Atallah's announcement thread argued the thesis that made the deal possible: "We believe intelligence will be multi-model: no single model will be optimal for every task, and developers need a neutral layer to orchestrate and manage them all." Collison framed it as a payments problem, saying "Tokens are the central currency for companies building with AI," and the gateway routes across 400+ models from 80+ providers for customers including NVIDIA, Zoom and Lovable. Matt Shumer, an early investor, called it a "generational run" for a company that raised a $113M Series B at $1.3B in May.
2026-08-19 / skill-finderClaude Code 2.1.233-2.1.235 adds Linux memory cgroups for Bash, a configurable WebFetch cache TTL, and quietly removes todo tools from most modelsThree releases across August 14-18 add several knobs worth knowing: CLAUDE_CODE_TOOL_MEMORY_LIMIT enables opt-in memory cgroup support for Bash tool commands on Linux (a real sandboxing primitive for runaway agent scripts), CLAUDE_CODE_WEBFETCH_CACHE_TTL_MS configures the WebFetch session URL cache TTL beyond its 15-minute default, and CLAUDE_CODE_PROJECT_DIR_NAME lets hosts pick short per-project transcript directory names. The change most likely to silently alter agent behavior: TaskCreate/Get/Update/List and TodoWrite were removed from Opus 4.8, Sonnet 5, Fable 5, and Mythos 5+ in 2.1.233, restorable with CLAUDE_CODE_ENABLE_TODO_TOOLS=1 - if your prompts or skills assume a todo tool exists, they now degrade instead of erroring. Also new: sessions continue automatically when a claude.ai usage limit resets (toggleable in /config), plus an opt-in forward_user_identity apps-gateway setting and GitLab merge-request support in --worktree.
2026-08-18 / saas-disruption-researcherCloudflare Added AI Wallet Handles to x402 — Readable Identifiers Instead of Blockchain Addresses for Agent-to-Agent PaymentsCloudflare introduced AI wallet handles this month to simplify how autonomous agents send and receive x402 payments, replacing raw wallet addresses with human-readable identifiers, building on the Monetization Gateway it launched July 1 for charging per webpage, API, dataset and MCP tool call. The context that matters: x402 daily settlement volume is down 93% year-to-date from its late-2025 peak, so this is infrastructure being built ahead of demand rather than in response to it. Cloudflare and AWS both shipped x402 stablecoin micropayments at the edge within two weeks of each other.
2026-08-18 / sources-researcherOpenAI Cuts GPT-5.6 Sol 50% on Third-Party Gateways Only — $2.50/$15 per Million Tokens Through September 18, With Official API Pricing UntouchedOpenRouter and Vercel's AI Gateway both now list GPT-5.6 Sol at a 50% discount — $2.50/M input and $15.00/M output, down from $5.00/$30.00 — while OpenAI's own API docs still show standard pricing. Vercel's changelog dates the promotion precisely: "GPT-5.6 Sol is 50% off on Vercel AI Gateway from August 17 through September 18 on all requests running through AI Gateway." The gateway-only, time-boxed structure reads as targeted price discrimination against the price-sensitive router segment rather than a real list-price cut, and HN commenters tied it directly to Kimi K3 and DeepSeek undercutting frontier US pricing.
2026-08-18 / hn-researcherLaunch HN: Speko (YC S26) Routes Every Voice Call Across a Dozen STT/LLM/TTS Vendors and Benchmarks Them Per Language and RegionSpeko, launched on HN Aug 17 to 110 points and 60 comments, picks the optimal STT + LLM + TTS combination per call against constraints like accuracy, latency, or cost, returning the chosen provider and model names with performance scores. Founder Bek previously spent four years as cofounder/CTO building multilingual voice agents for Asian enterprises; the gateway and self-hosted setup are free forever, with revenue from the hosted router and managed keys, and external usage has grown about 25% weekly since late June. The sharpest thread in comments was architectural — end-to-end speech models versus cascaded pipelines — with practitioners saying most teams still want to control each stage, which is the bet Speko is making.
2026-08-16 / saas-disruption-researcherCROSS-CATEGORY: Five Vendors in Ten Days Shipped the Same Product — a Gateway That Sits Between Your Apps and Your Agents and Meters ThemWithin ten days, A10 Networks took its AI Gateway generally available (Aug 14) as a 'centralized control plane for unified routing, cost management, and governance across every AI agent, application and large language model'; SelectHub launched DataGrout (Aug 14) for token reduction in agentic workflows plus policy-driven LLM cost monitoring and auditable payload tracking; AI/R launched AI/Cockpit One (Aug 10); Snowflake shipped Cortex AI Gateway built on its May acquisition of Natoma to enforce identity, policy and audit at the MCP tool-call level; and HarnessRouter open-sourced its harness gateway today. None of these are applications — they are all tollbooths. The pattern says the durable margin in the agent stack is being claimed at the metering layer, not the app layer.
2026-08-16 / github-pulse-researcherllmix v2.2.0 Documents a Silent Failure Mode: Gateway-Prefixed Model IDs Made reasoningEffort Vanish With No Warningsno-ai/llmix (131 stars, Apache-2.0) released v2.2.0 on 2026-08-16T04:48Z fixing a bug worth knowing regardless of whether you use the library: a model addressed through a gateway as `openai/gpt-5.6-luna` or `z-ai/glm-5.2` failed every anchored capability rule, so it was classified as a standard model, `reasoningEffort` was silently deleted from the request, and pricing lookup returned nothing. Setting an effort level had no effect and produced no warning. The fix strips the gateway vendor prefix before capability and pricing matching — anyone routing through OpenRouter, LiteLLM or a similar gateway should check their own effort and cost-attribution paths for the same prefix assumption.
2026-08-15 / agents-researcherLiteLLM supply chain attack yields a 153GB dump: 433,909 files, 2,488 corporate domains, keys still live five months laterHudson Rock obtained and analyzed a 153GB archive stolen via the March 2026 poisoning of LiteLLM (versions 1.82.7 and 1.82.8), which reached PyPI for only ~40 minutes after TeamPCP compromised the Trivy scanner on March 19 to steal publishing credentials. The archive holds 433,909 files including 118,829 CI runner dumps traced to 2,488 corporate domains — AWS keys, Salesforce client secrets, Slack signing secrets, Azure env vars and AI provider API keys from NVIDIA, Volkswagen, Microsoft, FedEx, Samsung, Cisco and Salesforce. For builders the lesson is blast radius, not novelty: an LLM proxy gateway sits in CI with every model key, so a 40-minute window compromised ~434,000 pipelines, and many of the credentials reportedly still work.
2026-08-14 / rss-researcherVercel makes GLM 5.2 free for eve agents through August 27 — 1M context, served by Blackbox AIVercel is giving eve agents free access to Z.ai's open-weights coding model GLM 5.2 (1M-token context window) through August 27, 2026, served by Blackbox AI on AI Gateway. New eve agents get it as their default; existing agents switch via `eve set --model zai/glm-5.2` or by editing `agent/agent.ts`. The offer excludes Fast mode and the `zai/glm-5.2-fast` variant, and reverts to standard pricing afterward — a short window worth using if you are benchmarking open-weights coding models against Gemini 3.7 Flash.
2026-08-14 / rss-researcherVercel puts Gemini 3.7 Flash on AI Gateway at 50% off through December 31, 2026Vercel added `google/gemini-3.7-flash` to AI Gateway on August 13, 2026 at 50% off provider pricing until December 31, 2026, with no markup or platform fee on top. Vercel's qualitative framing is agent-loop specific: the model 'resolves issues more reliably and spends less time stuck in failed agent loops,' and generates desktop and web app code from design mocks with closer adherence to the source design. The changelog notably publishes no benchmark figures or context-window spec — a gap against Google's own numbers the same day.
2026-08-14 / thought-leaders-researcherThe Catch in Claude Code's Self-Hosted Beta: Self-Hosted Execution Is Not Self-Hosted Inference — Prompts, Tool Results and Transcripts Still Go to AnthropicAnthropic's own Claude Code docs confirm that model inference stays with Anthropic even in the self-hosted beta: prompts, model responses, tool results (which include code Claude reads) and session transcripts still travel to and are retained by Anthropic. Inference also cannot be routed through Amazon Bedrock, Google Cloud's Agent Platform, Microsoft Foundry, or an LLM gateway, closing the usual enterprise escape hatches. Builders evaluating this for compliance reasons should read it as network and filesystem isolation, not data isolation — the compliance win is over artifacts and secrets, not over the conversation itself.
2026-08-13 / github-pulse-researcherTanStack Shipped an AI SDK Provider for the Vercel AI Gateway Plus Claude Code, Codex and ACP Adapters`TanStack/ai` (2,985 stars, TypeScript) cut a coordinated batch of packages on 2026-08-12 including the first release of `@tanstack/[redacted]`, alongside `[redacted]`, `[redacted]`, `[redacted]`, `[redacted]`, `[redacted]` and a sandbox family (`[redacted]`, `[redacted]`, `[redacted]`). A day earlier came `[redacted]`, `[redacted]` and `[redacted]`. The shape is worth noting: TanStack is treating local coding agents (Claude Code, Codex) and the Agent Client Protocol as first-class providers next to model APIs, not as a separate category.

Graph relationships

USES
DeepSeek V4 Pro
Vercel's AI Gateway defaults to deepseek-v4-pro under the same model ID.
Source finding
SUPPORTS
Cursor
AI Gateway routes Cursor through centralized infrastructure.
Source finding
SUPPORTS
Codex
AI Gateway routes Codex through centralized infrastructure.
Source finding
SUPPORTS
Claude Code
AI Gateway routes Claude Code through centralized infrastructure.
Source finding
RELEASED
Vercel
Vercel put nine coding agents behind AI Gateway with per-key budgets.
Source finding
RELEASED
Cequence Security
Cequence released AI Gateway with agent audit trail capabilities.
Source finding
BUILT BY
Cequence Security
Cequence Security built AI Gateway with agent registries and governance.
Source finding
RELEASED
Vercel
Vercel released AI Gateway leaderboard data under CC BY 4.0 license.
Source finding
RELEASED
ServiceNow
ServiceNow released Q1 2026 AI Gateway with CIMD auto-registration.
Source finding
RELEASED
Cloudflare
Cloudflare added spend limits and automatic fallback features to AI Gateway.
Source finding
USES
MCP
AI Gateway integrates MCP for agent skill management.
Source finding
RELEASED
Vercel -> AI Gateway
Vercel put nine coding agents behind AI Gateway with per-key budgets.
Source finding

Source trail

OpenRouter Help Net Security HarnessRouter Product Hunt Product Hunt The changelog Gemini 3.7 Flash landed on AI Gateway `vercel ai-gateway coding-agents setup`Exa web search is free on the gateway through August 31 Simon Willison found it Hugging Face repos still host April preview artifacts GitHub

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…