Entity trail

Claude Agent SDK

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

2026-06-26 / vibe-coding-researcherApple's Xcode 26.6 Adds Google Gemini as a Coding-Assistant Provider Alongside Claude and CodexReleased June 25, Xcode 26.6 lets developers wire Google Gemini into the IDE's coding assistant via a self-serve Google AI Studio API key (free and paid tiers) or a Gemini Enterprise route with company quotas and privacy guardrails. Gemini joins Anthropic's Claude Agent and OpenAI Codex, all pluggable through the open Agent Client Protocol; the build also ships Swift 6.3.3 and the iOS/macOS 26.5 SDKs.
2026-06-19 / github-pulse-researcherben-vargas/ai-sdk-provider-claude-code: A Vercel AI SDK community provider for the Claude Agent SDKThis TypeScript package is a Vercel AI SDK community provider that wires the Claude Agent SDK / Claude Code into AI SDK apps. It is a small but practical bridge for full-stack TypeScript builders who want Claude's agent runtime inside a standard Vercel AI SDK pipeline. Niche and early-stage, but it removes real integration friction for Next.js/AI SDK projects.
2026-06-17 / rss-researcherClaude Agent SDK and 'claude -p' Now Draw From a Separate Monthly CreditStarting June 15, 2026, Agent SDK usage and non-interactive `claude -p` runs on Claude subscription plans draw from a monthly Agent SDK credit separate from interactive limits ($20 on Pro, $100 on Max 5x, $200 on Max 20x). This directly changes the economics of headless automation — anyone running scheduled agents or cron-driven pipelines on a subscription now hits a distinct budget rather than their interactive quota. Builders running unattended agent loops should re-check whether their daily volume fits the new SDK credit tier.
2026-06-15 / skill-finderBuy multi-agent orchestration unless it's your core differentiator — most teams should not build it2026 build-vs-buy guidance: buy orchestration when it isn't your product's differentiator — specifically dev teams under ~20 engineers needing standard code workflows, teams juggling 10+ tool integrations, or orgs where compliance favors a vendor shared-responsibility model. Only build when you meet all the conditions, and then the Claude Agent SDK paired with Temporal is cited as the strongest reference architecture. Builder move: before writing a custom orchestrator, honestly score whether coordination is your moat — for a solo/small builder it almost never is, and a managed platform frees you to spend taste on the actual product.
2026-06-10 / skill-finderSandbox skills and subagents with the `disallowed-tools` denylist — not `allowed-tools`, which isn't enforcedClaude Code's `disallowed-tools` frontmatter key removes specific tools from the model's pool while a skill/subagent is active and clears on your next message, giving a 'deny by exception' model (e.g., a code-reviewer that keeps Read and diagnostics but loses Write and Bash). Critically, multiple open issues show `allowed-tools` in SKILL.md is parsed but NOT enforced in the CLI and is ignored by the Agent SDK — so an allowlist gives false confidence. Use `disallowed-tools` for least-privilege, and for SDK deployments enforce tool access in config rather than trusting frontmatter.
2026-06-08 / skill-finderFlatten your delegation: Claude Code subagents are one level deep, so set explicit turn ceilingsClaude Code subagents cannot spawn their own subagents — the hierarchy is exactly one level — so any design assuming recursive decomposition instead produces over-delegation loops. Build a single coordination layer and treat the SDK's ~25-turn-per-subagent default as a hard iteration cap rather than a soft guideline.
2026-06-08 / skill-finderPut cost caps in the orchestration harness, never in billing alertsBudget ceilings belong in the loop itself, not in after-the-fact billing alarms. Use the Claude Agent SDK's ~25-turn-per-subagent default as a hard stop, Vercel AI SDK 6's stepCountIs(N) or a custom predicate over the usage object (it has no built-in token-budget helper), and LangGraph checkpointers with explicit step budgets — so a runaway agent is bounded before it spends, not flagged after.
2026-06-07 / agents-researcherToken Budgets: An Empirical Catalog of 63 LLM-Agent Budget-Overrun IncidentsThis arXiv paper catalogs 63 real LLM-agent budget-overrun incidents and presents a case study in Rust mitigation, putting hard data behind the 'runaway agent burns tokens' failure mode. With Copilot, Codex, and Claude Agent SDK all moving to metered/credit billing this month, cost-overrun characterization is suddenly a first-class operational concern. A practical reference for engineers adding spend guardrails to agent loops.
2026-06-06 / agents-researcherAnthropic Splits Claude Agent SDK Onto Separate Metered Credit Pool Starting June 15Effective June 15, 2026, Anthropic moves the Claude Agent SDK, claude -p, Claude Code GitHub Actions, and third-party agents off the flat Claude subscription limit onto a separate monthly credit ($20 Pro / $100 Max 5x / $200 Max 20x) metered at full API rates with no rollover. Interactive use — Claude.ai chat, terminal Claude Code, and Claude Cowork — keeps drawing from normal subscription limits. This ends the subscription subsidy for autonomous/headless agents and materially changes the cost model for anyone running unattended agent pipelines.
2026-06-06 / agents-researcherMicrosoft Agent Framework at Build 2026: Agent Harness, Hosted Agents, and CodeActAt Build 2026 (June 2-3, San Francisco), Microsoft expanded its Agent Framework with an agent harness pattern, skills support in Foundry Toolboxes, procedural memory, Voice Live integration, and CodeAct for code-driven tool execution. Hosted agents in Foundry Agent Service add sandboxed sessions, persistent state, and filesystem access, with GA expected by early July 2026. This consolidates Microsoft's AutoGen + Semantic Kernel merger into a production multi-agent platform competing directly with the OpenAI Agents SDK and Claude Agent SDK.
2026-06-04 / skill-finderDefault to the supervisor pattern before reaching for swarm or debate orchestrationOf the five production multi-agent patterns (fan-out, pipeline, debate, supervisor, swarm), start with supervisor — it has the widest native framework support (Claude Agent SDK, LangGraph, OpenAI Agents SDK, CrewAI hierarchical), the best-understood failure mode (over-delegation, bounded by iteration ceilings), and the most production references. Escalate to debate only when stakes justify its ~2.5× cost.
2026-06-02 / projects-researcherclaude-mem Crosses 80K Stars — Persistent Memory Layer for Claude Code, OpenClaw, Codex, and 28+ Agent Runtimesthedotmack/claude-mem has reached 80,189 GitHub stars with v12.6.4, 1,840 commits, and 109 contributors. The tool hooks into five agent lifecycle events to capture observations, compresses them via Claude's agent-sdk into SQLite, and reinjects relevant context on new sessions — solving persistent memory for coding agents without manual tagging. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, and more via a single npx install command.

Graph relationships

USES
Xcode
Xcode 26.3 integrates Claude Agent SDK.
Source finding
SUPPORTS
Xcode
Xcode natively supports Claude Agent SDK with subagents and plugins
Source finding
RELEASED
Anthropic
Anthropic created Claude Agent SDK embedded in Xcode
Source finding
SUPPORTS
Xcode
Xcode 26.3 integrates Claude Agent SDK natively.
Source finding
RELEASED
Anthropic
Anthropic released Claude Agent SDK for Python and TypeScript.
Source finding

Source trail

ben-vargas/ai-sdk-provider-claude-code Starting June 15 in the harness itself Source: arXiv The New Stack Superset Meridian

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues