# Agent Infrastructure Published articles for Agent Infrastructure. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## AI agent credentials: Four architectures compared DevFeed: [AI agent credentials: Four architectures compared]() Original publisher: [Read original article]() Author: WorkOS Published: 2026-09-21T00:00:00Z Content type: comparison Language: en Sources: [WorkOS]() Topics: [AI Agent](), [Claude](), [Google](), [Nvidia](), [OpenShell](), [proxy](), [Language models](), [Network](), [OAuth 2.0](), [Model Context Protocol]() Tags: [agent](), [agent-infrastructure](), [ai](), [ai-agent](), [architectures](), [credentials](), [google](), [mcp](), [network](), [nvidia](), [oauth2](), [production](), [proxy](), [secrets]() ### AI overview A comparison of four approaches to securing credentials for production AI agents. Google and NVIDIA keep secrets outside the sandbox, Rubrik issues tokens per tool call, Microsoft inspects MCP traffic, and Opal delegates request enforcement. The article focuses on which failures each architecture prevents and the gaps that remain. ### Source excerpt Google, NVIDIA, Rubrik, Microsoft and Opal each solve agent credentials differently. What each one protects against, and the gap they share. ## OpenAI's Model-Misalignment Reporting Framework Offers an Incident Template for Agent Teams DevFeed: [OpenAI's Model-Misalignment Reporting Framework Offers an Incident Template for Agent Teams]() Original publisher: [Read original article]() Author: Developers Digest Published: 2026-09-19T00:00:00Z Content type: opinion Language: en Sources: [Developers Digest]() Topics: [OpenAI](), [incident](), [Template](), [Tool]() Tags: [agent](), [agent-infrastructure](), [agents](), [ai-agents](), [ai-safety](), [developer-tools](), [disclosure](), [incident](), [openai](), [reporting](), [signal](), [template]() ### AI overview The document presents OpenAI's model-misalignment reporting framework as a practical template for teams operating tool-using agents, covering incident intake, severity labels, and evidence-led disclosure. ### Source excerpt OpenAI's model-misalignment reporting framework is not just a safety-policy document. For teams shipping tool-using agents, it is a template for incident intake, severity labels, and evidence-led disclosure. ## Claude Code Plugin Evals Make Agent Extensions Testable DevFeed: [Claude Code Plugin Evals Make Agent Extensions Testable]() Original publisher: [Read original article]() Author: Developers Digest Published: 2026-09-13T00:00:00Z Content type: release Language: en Sources: [Developers Digest]() Topics: [Claude Code](), [Code]() Tags: [agent](), [agent-infrastructure](), [agent-skills](), [ai-coding](), [ci](), [claude-code](), [developer-tools](), [evals](), [html](), [json](), [plugin](), [reports]() ### AI overview Claude Code 2.1.269 introduces plugin evaluations, baseline comparisons, JSON and HTML reporting, and CI gates, making agent extensions measurable. ### Source excerpt Claude Code 2.1.269 adds plugin evals, baseline comparisons, JSON and HTML reports, and CI gates. That changes plugins from clever prompts into measurable agent infrastructure. ## Hermes Agent Adds Vercel AI Gateway and Sandbox Backends DevFeed: [Hermes Agent Adds Vercel AI Gateway and Sandbox Backends]() Original publisher: [Read original article]() Author: Developers Digest Published: 2026-08-08T00:00:00Z Content type: news Language: en Sources: [Developers Digest]() Topics: [Artificial Intelligence](), [Vercel](), [Model Routing](), [Open Source](), [gateway](), [Terminal](), [Cloud](), [control-plane](), [Back end]() Tags: [agent-infrastructure](), [ai](), [ai-gateway](), [cloud](), [coding-agents](), [control-plane](), [model-routing](), [news](), [open-source](), [terminal](), [vercel]() ### AI overview Vercel added Hermes Agent to its AI Gateway and made Vercel Sandbox a terminal backend for the open-source agent. The setup supports user-provided model routing across more than 200 models without markup and a cloud microVM for each agent command. ### Source excerpt Vercel added Hermes Agent to AI Gateway and made Vercel Sandbox a terminal backend for the open-source agent. Hermes is now fully BYO: your own model routing through 200+ models at no markup, and your own cloud microVM for every agent command. Here is what that unlocks and why the agent control plane is consolidating. ## Cloudflare Folds Workers AI Into AI Gateway: One Control Plane for Every Model Provider DevFeed: [Cloudflare Folds Workers AI Into AI Gateway: One Control Plane for Every Model Provider]() Original publisher: [Read original article]() Author: Developers Digest Published: 2026-08-07T00:00:00Z Content type: release Language: en Sources: [Developers Digest]() Topics: [Cloudflare](), [control-plane](), [Artificial Intelligence](), [API](), [REST API](), [Routing (disambiguation)]() Tags: [agent-infrastructure](), [ai](), [ai-gateway](), [api](), [cloudflare](), [control-plane](), [gateway](), [model-routing](), [news](), [rest-api](), [routing]() ### AI overview Cloudflare is merging Workers AI and AI Gateway into a unified control plane with a shared /ai/ REST API, automatically created default gateways, transferable AI Gateway credits for Workers AI, and model-first provider routing. ### Source excerpt Cloudflare is merging Workers AI and AI Gateway into one control plane: unified /ai/ REST API, auto-created default gateways, AI Gateway credits spendable on Workers AI, and model-first routing that picks the provider for you. Here is what changes and what stays. ## Chat SDK Adds Durable Approvals: Agent Workflows That Wait For a Human DevFeed: [Chat SDK Adds Durable Approvals: Agent Workflows That Wait For a Human]() Original publisher: [Read original article]() Author: Developers Digest Published: 2026-08-06T00:00:00Z Content type: release Language: en Sources: [Developers Digest]() Topics: [SDK](), [Vercel]() Tags: [agent](), [agent-infrastructure](), [ai-agents](), [chat](), [clicks](), [developer-workflow](), [handler](), [news](), [sdk](), [table](), [thread](), [vercel](), [workflow](), [workflows]() ### AI overview Vercel's Chat SDK adds durable approval controls for agent workflows. A workflow can pause until a person approves an action in a chat thread, with the wait surviving deployments and decisions scoped to authorized approvers. ### Source excerpt Vercel's Chat SDK can now suspend a Workflow SDK run until someone clicks Approve in a chat thread. One requestApproval call replaces the approvals table, the onAction handler, and the polling loop - with verified decisions, scoped approvers, and a wait that survives deploys. ## DeepSeek V4 Flash Is 90% Off Through Novita on Vercel AI Gateway: The Cost Math DevFeed: [DeepSeek V4 Flash Is 90% Off Through Novita on Vercel AI Gateway: The Cost Math]() Original publisher: [Read original article]() Author: Developers Digest Published: 2026-08-05T00:00:00Z Content type: news Language: en Sources: [Developers Digest]() Topics: [Artificial Intelligence](), [deepseek](), [gateway](), [Vercel]() Tags: [agent](), [agent-infrastructure](), [ai](), [ai-gateway](), [cost](), [deepseek](), [effective](), [flash](), [gateway](), [input](), [news](), [pricing](), [routing](), [setup](), [tokens](), [vercel]() ### AI overview DeepSeek V4 Flash is temporarily available through Novita on Vercel AI Gateway at a 90% discount for Pro customers, reducing the effective input and output rates to $0.014 and $0.028 per million tokens. The article presents the before-and-after cost calculation, provider-pinning setup, and implications for routing decisions in inexpensive agent loops. ### Source excerpt DeepSeek V4 Flash routed to Novita on Vercel AI Gateway is 90% off for Pro customers through August 11, dropping the effective rate to $0.014 input / $0.028 output per million tokens. Here is the verified before/after math, the provider-pinning setup, and what a 10x cheap agent loop means for routing decisions. ## The v0 API Is GA: Vercel Just Made Its App-Building Agent a Headless Service DevFeed: [The v0 API Is GA: Vercel Just Made Its App-Building Agent a Headless Service]() Original publisher: [Read original article]() Author: Developers Digest Published: 2026-08-05T00:00:00Z Content type: release Language: en Sources: [Developers Digest]() Topics: [API](), [App](), [Vercel](), [service](), [prompt](), [async](), [Streaming]() Tags: [agent](), [agent-infrastructure](), [ai-agents](), [ai-coding](), [api](), [app](), [async](), [building](), [deploy](), [headless](), [live](), [news](), [preview](), [prompt](), [running](), [service](), [streaming](), [vercel]() ### AI overview The v0 API is generally available, providing programmatic, headless access to v0's app-building agent. Developers can send a prompt, receive a running app with an embeddable live preview URL, and deploy it to Vercel in one call. The release also describes synchronous, asynchronous, and streaming operation in agent workflows. ### Source excerpt The v0 API is now generally available: programmatic, headless access to v0's app-building agent. Send a prompt, get a running app with a live preview URL you can embed, then deploy to Vercel in one call. Here is what changed, how the sync/async/streaming model works, and how it fits in an agent loop. ## Vercel AI Gateway Adds Team and Project Spend Budgets: The Cost-Cap Math for Agent Builders DevFeed: [Vercel AI Gateway Adds Team and Project Spend Budgets: The Cost-Cap Math for Agent Builders]() Original publisher: [Read original article]() Author: Developers Digest Published: 2026-08-01T00:00:00Z Content type: release Language: en Sources: [Developers Digest]() Topics: [Artificial Intelligence](), [gateway](), [Vercel](), [Command-line interface]() Tags: [agent-infrastructure](), [ai](), [ai-gateway](), [cli](), [cost](), [cost-control](), [gateway](), [limits](), [news](), [projects](), [team](), [vercel]() ### AI overview Vercel AI Gateway introduces spend budgets scoped to teams and projects. Hard dollar limits can reject requests, while email alerts trigger at 50%, 75%, and 100%; defaults can be managed through the CLI. ### Source excerpt AI Gateway spend budgets now scope to teams and projects, with hard dollar limits that reject requests, email alerts at 50/75/100%, and CLI-managed defaults. Here is how the three scopes compose and where it fits your cost stack. ## Tencent Hunyuan Paper Maps Coding-Agent Behavior Across Prompts, Tools, State, Permissions, and Runtime Policy DevFeed: [Tencent Hunyuan Paper Maps Coding-Agent Behavior Across Prompts, Tools, State, Permissions, and Runtime Policy]() Original publisher: [Read original article]() Author: Developers Digest Published: 2026-07-16T00:00:00Z Content type: article Language: en Sources: [Developers Digest]() Topics: [code search](), [Tool](), [Authorization](), [Prompt Engineering]() Tags: [2026](), [agent](), [agent-infrastructure](), [ai-agents](), [code-search](), [codex](), [developer-workflow](), [evals](), [harness](), [permissions](), [runtime](), [search](), [tencent]() ### AI overview A July 2026 Tencent Hunyuan paper maps coding-agent behavior across prompts, tools, state, permissions, and runtime policy, arguing that code search alone is insufficient. ### Source excerpt A July 2026 paper from Tencent Hunyuan turns agent harnesses into behavior-level maps. The useful lesson for builders is simple: code search is not enough when one behavior spans prompts, tools, state, permissions, and runtime policy. ## Image Token Compression Is a Real Agent Cost Lever DevFeed: [Image Token Compression Is a Real Agent Cost Lever]() Original publisher: [Read original article]() Author: Developers Digest Published: 2026-07-04T00:00:00Z Content type: article Language: en Sources: [Developers Digest]() Topics: [Compression](), [context]() Tags: [accounting](), [agent](), [agent-infrastructure](), [ai-coding](), [claude-code](), [compression](), [context](), [cost](), [cost-optimization](), [evals]() ### AI overview A Show HN project claims that rendering bulky agent context as images can reduce costs. The document emphasizes evaluating compression, enforcing byte-safety rules, and accounting for each request. ### Source excerpt A Show HN project claims large agent-cost cuts by rendering bulky context as images. The useful lesson is not the trick itself. It is that compression needs evals, byte-safety rules, and per-request accounting.