AgentStackWatch
⌕Subscribe

Issue No. 0 · October 17, 2026

The AI agent stack,
tracked every night.

An independent publication on personal agents, managed platforms, frameworks and the research behind them — with a live tracker of 43 products and every fact sourced.

Free: The Personal AI Agent Buyer's Guide →For founders and execs. Every fact sourced.

43 products23 source-verified20 repo-verified1.4M GitHub stars tracked21 release feeds
Latest releases
OpenClaw 2026.10.4Oct 16LangGraph 1.2.16Oct 16Pydantic AI 2.57.0Oct 16Mastra 1.78.0Oct 16Vercel AI SDK 7.0.143Oct 16Goose 1.55.0Oct 16Hermes Agent 0.22.2Oct 15OpenAI Agents SDK 0.24.0Oct 15Claude Agent SDK 0.2.168Oct 15Haystack 3.4.0Oct 15Browser Use 0.13.11Oct 15CrewAI 1.15.28Oct 14E2B 2.56.0Oct 13Google ADK 2.12.0Oct 13DSPy 3.4.1Oct 12OpenHands 1.27.1Oct 12

October 16, 2026

What actually changed in AI agents this week

Issue #1, Oct 12–16, 2026: everyone shipped context compaction and human approvals, two personal agents shipped stability patches, and the download charts tell a different story than the hype.

Read the issue →

Agent Stack Weekly · free

One email a week. Everything that shipped in AI agents.

Releases, pricing changes and breaking changes from the tracker, plus a short take from a working agent architect.

Form not loading? Subscribe on Beehiiv →

The landscape

Who's building the agent stack

Open full map →
PERSONAL AGENTSMANAGED AGENT PLATFORMSDEV FRAMEWORKS & SDKSRESEARCH & BENCHMARKS OpenClaw Hermes Agent Daytona Browser Use OpenHands CrewAI Goose LlamaIndex LangGraph DSPy OpenAI Agents SDK smolagents Mastra Vercel AI SDK Haystack Letta Google ADK
Circle area ∝ GitHub stars (fetched 2026-10-17). Hollow circles are closed-source products with no public repo. Hover for details, click to open.

The tracker

Four corners of the stack

Browse all 43 →

Personal agents

Consumer and self-hosted assistants that act on your behalf.

10 entries

Managed agent platforms

Hosted runtimes, agent APIs and sandboxes.

12 entries

Dev frameworks & SDKs

Libraries for building agents in code.

16 entries

Research & benchmarks

Leaderboards and benchmarks used to evaluate agents.

5 entries