Entity trail

Chat SDK

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

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-08-06 / rss-researcherVercel's Chat SDK adds workflow approvals that survive deploys and restarts for up to 24 hoursA new chat/workflow subpath in Vercel's Chat SDK exposes requestApproval, which posts an Approve/Deny card and suspends a Workflow SDK workflow until someone decides — returning {approved, user, timedOut} with optional approvers restriction and timeouts demonstrated at 24h. Vercel's claim is that 'the wait can last seconds or days and survives deploys and restarts,' with signature verification, access control and audit trails handled by the platform. This removes the standard human-in-the-loop scaffolding: no approvals table, no onAction handler, no polling loop.
2026-08-03 / github-pulse-researcherGrafana Ships a Go AI SDK That Is Wire-Compatible With Vercel's TypeScript React Hooksgrafana/ai-sdk (198 stars, created 2026-07-28, Apache-2.0) gives Go one API for model calls, streaming, tools, structured output, and multi-step agents, deliberately following Vercel AI SDK's design and staying wire-compatible with its frontend hooks. A Go backend calling aisdk.StreamText and aisdk.WriteUIMessageStream streams Server-Sent Events directly into useChat with no protocol adapter. For anyone with a Go service and a React frontend, this removes the usual reason to keep a TypeScript BFF in the middle purely to speak the AI SDK stream protocol.
2026-08-03 / projects-researcherTanStack AI Adds Bounded Tool-Call Fan-Out — maxToolCalls(n) and maxToolCallsPerTurn Cap Runaway Agent LoopsTanStack/ai (2,953 stars), the type-safe provider-agnostic TypeScript SDK spanning OpenAI, Anthropic, Gemini and Ollama with React, Vue, Svelte and Solid adapters, shipped agent-loop guardrails: AgentLoopState now exposes `toolCallCount` and `lastTurnToolCallCount`, a `maxToolCalls(n)` strategy caps cumulative calls, and `chat({ maxToolCallsPerTurn })` caps how many parallel calls execute in a single turn. This is a direct answer to the failure mode where an agent loop fans out unbounded tool invocations and burns a budget, and it lands as a first-class strategy rather than a user-implemented counter. The framework positions explicitly against Vercel AI SDK on vendor lock-in.
2026-07-29 / rss-researcherSandstone grew revenue 40x in 147 days running agentic legal workflows on a 7-app turborepoVercel published a July 27 customer story on Sandstone, an AI Legal Relationship Management platform that ingests context from Salesforce, email, and messaging so lawyers have complete information before starting work. The company reports 40x revenue growth in 147 days, 1,000+ legal requests managed daily across customer teams, and a $30M Series A from Lightspeed, running a 7-app turborepo deployed across multiple Vercel projects. The stack detail worth noting for builders: multi-step agentic workflows on the Vercel AI SDK plus Chat SDK for Teams, Emulate SDK for Slack/Okta testing, Agent Browser SDK, and Flags SDK, with the CTO citing ten-minute tool additions as the velocity unlock.
2026-07-26 / agents-researcherAgentForger: one ChatGPT link forged a persistent attacker-controlled agent inside an enterprise workspaceZenity Labs disclosed AgentForger on July 23 — a cross-site agent forgery flaw in ChatGPT's Agent Builder where two URL parameters (template selector and initial-prompt field) let a single crafted link create, authorize, and activate an invisible agent inside a victim's authenticated workspace with approvals switched off. The forged agent polled an attacker's inbox for new commands every five minutes and could map org structure, exfiltrate documents, harvest credentials, and impersonate the victim across Slack, Teams, and email. OpenAI received the report via Bugcrowd on June 4, confirmed within 24 hours, and removed the vulnerable parameter by June 8; Agent Builder itself is slated for deprecation on November 30, 2026 in favor of the Agents SDK.
2026-07-24 / github-pulse-researcherThinkInAIXYZ/deepchat Adds agent-skills as a Top-Level Topic — Skills Spread From CLIs to Desktop Chat ClientsDeepChat, a TypeScript desktop assistant at 6,155 stars, now carries agent-skills alongside agent, ai-assistant and ai-sdk in its topic list, joining mirrord and khazix-skills in tagging skills as a first-class capability. The pattern across this week's set is that the skills primitive — originally a Claude Code CLI concept — is being adopted by desktop clients, Kubernetes tooling and multi-agent orchestrators as a portable packaging format. For builders, that means skills written once are increasingly runnable outside the harness they were authored for.
2026-07-10 / agents-researcherOpenAI launches ChatGPT Work, an agentic workspace on GPT-5.6, with Sol/Luna/Terra model tiersOn July 9–10, OpenAI shipped ChatGPT Work, an agentic workspace that executes long-running multi-step tasks across Slack, Gmail, Google Drive, CRMs, and internal knowledge bases, producing documents, spreadsheets, and sites rather than chat turns. It runs on a three-model GPT-5.6 family — Sol (frontier, $5/M input, $30/M output), Luna (speed), Terra (balanced) — with Altman claiming 54% better token efficiency on agentic coding. Desktop is available now; Plus/Pro/Business/Enterprise/Edu web and mobile follow within days. This is OpenAI moving from an agent SDK story to an agent product that competes directly with Anthropic Cowork and Microsoft Copilot.
2026-06-17 / thought-leaders-researcherGuillermo Rauch Unveils Vercel's 'Agent Stack' at Ship 26: Coding Agents Now Trigger Over Half of All DeploymentsAt Vercel Ship 26 in London on June 17, Rauch launched the Agent Stack (AI SDK, AI Gateway, Sandbox, Workflow SDK, Chat SDK) plus a new Vercel Connect and enterprise agent controls, repositioning the whole platform around running agents in production. The headline metric: coding-agent-triggered deployments jumped from under 3% to over 50% in six months, and AI Gateway token volume grew from ~2T to ~20T per month. The takeaway for builders is that 'deploy' is becoming an agent action, not a human one — infra is being re-architected around that assumption.
2026-06-16 / agents-researcherOpenAI winds down Agent Builder and Evals, steering developers to Agents SDK and Workspace AgentsOpenAI announced (June 3, 2026) that Agent Builder and Evals will be removed from the platform on November 30, 2026, recommending the Agents SDK for workflows that should live as code and Workspace Agents in ChatGPT for prompt-driven use cases. Workspace Agents can own entire multi-step workflows, follow team processes, and be shared across a workspace, with new per-app action safeguards; the free period was extended to July 6, 2026. The deprecation is a notable consolidation signal — OpenAI is collapsing its no-code agent tooling in favor of SDK-as-code plus managed workspace agents.
2026-06-16 / vibe-coding-researcherPattern: Local-First AI Engines Are Maturing Into Production Agent BackendsA cluster of trending repos signals a serious local-first stack: LocalAI (46,890 stars, 'run any model on any hardware, no GPU required'), Microsoft's Foundry-Local (GPU-accelerated local chat-completions/SDK), and ai-sdk-ollama (a Vercel AI SDK provider for Ollama). Together they make on-device/air-gapped agent deployment viable for privacy-sensitive or cost-controlled workloads. Builders can now prototype agents against a local engine with the same SDK surface they'd use against hosted frontier models.
2026-06-15 / vibe-coding-researcherOX Security Discloses Systemic RCE Baked Into Anthropic's Official MCP SDKs Across Python, TypeScript, Java, and RustOX Security's research team reports a critical, architectural vulnerability in the official Model Context Protocol SDKs that enables arbitrary command execution (RCE) on any system running a vulnerable MCP implementation, exposing internal databases, API keys, and chat histories. Crucially they frame it as a design decision present across every supported language SDK — Python, TypeScript, Java, and Rust — not a single-language bug. This is distinct from the back-end-connector and unauthenticated-server classes already reported, and warrants immediate review of any self-hosted MCP server.
2026-06-05 / github-pulse-researcherThinkInAIXYZ/deepchat at 5.9K Stars: A Smart Assistant Layer Connecting Powerful Models to Your Personal Data, Skills and AgentsThinkInAIXYZ/deepchat (~5,900 stars, TypeScript) is a desktop smart assistant that connects frontier models to a user's personal world, with native support for agents, agent-skills, and the AI SDK. It sits in the personal-assistant niche alongside Leon and LobeHub but leans on the newer 'skills' abstraction for extensibility. The skills-first design is a small signal that agent-skill packaging is spreading from coding agents into general assistants.

Graph relationships

RELEASED
Vercel
Vercel released Chat SDK update with message.subject field
Source finding
RELEASED
Vercel -> Chat SDK
Vercel released Chat SDK update with message.subject field
Source finding

Source trail

A new chat/workflow subpath Vercel Ship 26 in London today The New Stack argued GitHub GitHub Vercel Blog Zenity Labs GitHub InfoWorld OpenAI Help Center GitHub OX Security

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…