AutoGen maintenance puts your agent framework on notice

AutoGen has entered maintenance mode: Microsoft says no new features or enhancements are planned, and community management will take over. At the same time, Cognition has renamed Windsurf to Devin Desktop and positioned it as a control surface for local and cloud coding agents, while Claude Code outbound-click interest rose 35% week over week. The coding-agent stack is separating into framework dependencies, agent control surfaces, and execution tools. If AutoGen is in your stack, inventory its dependencies and assess your migration options alongside how you plan to govern local and cloud agents.

See what changed on AutoGen's listing →

What changed on the ground

Changes we detected on each tool's own pages — pricing, features, platform reach — via our daily crawls. Often ahead of any press coverage; each links to the scored listing.

What's new in the directory

Biggest movers this week

Listings with the largest week-over-week jump in outbound-click interest

• Claude Code · +35% week-over-week. Claude Code is Anthropic's agentic coding tool that lives in your terminal. It understands your entire codebase, makes multi-file edits, runs commands, manages git workflows, and uses MCP for tool integration. Built with a Unix philosophy — it reads, plans, edits, and verifies in a loop. The fastest-growing product in the coding agent category.
• Cline · +10.9% week-over-week. Cline is an AI coding assistant for VS Code that can inspect your project, edit files, run terminal commands, and use a browser while asking for permission at each step. It is aimed at developers working on real codebases who want more than code completion.
• Cursor · +1% week-over-week. Cursor is a developer-focused AI environment that adds agents, context, and automation around your repositories. It combines an editor-like interface, a CLI, and a cloud agent API to automate code review, bug fixing, CI hygiene, and more. Designed for individual developers and engineering teams who want AI to take real actions in their code and infrastructure, not just chat.

Biggest declines this week

Listings with the largest week-over-week drop in outbound-click interest

• Codex CLI · -56.2% week-over-week. Codex CLI is OpenAI’s coding agent that runs on your computer from the terminal. It can also be used in VS Code, Cursor, Windsurf, or as a desktop app.
• OpenClaw · -30% week-over-week. OpenClaw is a personal AI assistant that goes beyond chat and can run tasks, remember context, and act across connected tools. It is aimed at people who want an agent they can message like a coworker and let keep working in the background.
• OpenHands · -29.2% week-over-week. OpenHands (formerly OpenDevin) is a fully autonomous open-source AI coding agent. MIT-licensed core, plans + executes Python actions via CodeAct, uses browser + terminal + files. ~72% on SWE-bench Verified with Sonnet 4.5 via the OpenHands Software Agent SDK (arXiv:2511.03690).
• Hyper · -23.5% week-over-week. Hyper is a cloud-based autonomous software engineering platform for turning product objectives into tested code in your GitHub repositories. It plans work, runs parallel agent sessions in isolated lanes, verifies results server-side, and can open pull requests or deploy supported projects.
• Sinatra · -20.4% week-over-week. Sinatra is an autonomous coding agent that takes work from Linear or GitHub, reads the relevant repository, writes a change, runs tests, and opens a draft pull request. Teams review the result in GitHub and can request follow-up changes in the PR thread.

Agenticness Leaderboard

This week's most-clicked tools, re-ranked by rubric score. The score is the editorial signal; the popularity is what drew the directory's traffic here this week.

• Hyper · 22/36 · Domain Specialist · L3

Hyper is a cloud-based autonomous software engineering platform for turning product objectives into tested code in your GitHub repositories. It plans work, runs parallel agent sessions in isolated lanes, verifies results server-side, and can open pull requests or deploy supported projects.

• Sinatra · 21/36 · Domain Specialist · L3

Sinatra is an autonomous coding agent that takes work from Linear or GitHub, reads the relevant repository, writes a change, runs tests, and opens a draft pull request. Teams review the result in GitHub and can request follow-up changes in the PR thread.

• Claude Code · 20/36 · Domain Specialist · L3

Claude Code is Anthropic's agentic coding tool that lives in your terminal. It understands your entire codebase, makes multi-file edits, runs commands, manages git workflows, and uses MCP for tool integration. Built with a Unix philosophy — it reads, plans, edits, and verifies in a loop. The fastest-growing product in the coding agent category.

• Cursor · 18/36 · Adaptive Collaborator · L2

Cursor is a developer-focused AI environment that adds agents, context, and automation around your repositories. It combines an editor-like interface, a CLI, and a cloud agent API to automate code review, bug fixing, CI hygiene, and more. Designed for individual developers and engineering teams who want AI to take real actions in their code and infrastructure, not just chat.

• Cline · 18/36 · Adaptive Collaborator · L2

Cline is an AI coding assistant for VS Code that can inspect your project, edit files, run terminal commands, and use a browser while asking for permission at each step. It is aimed at developers working on real codebases who want more than code completion.

Learn: How AUX actually works

AUX doesn’t plan agent actions or take them for you. It is a cloud API that checks a business counterparty before a higher-stakes system acts. In plain terms, another agent or app sends AUX a company name or identifier, and AUX pulls bounded evidence from sources like GLEIF, OFAC, IANA RDAP, and any private attestations a user has configured. It then returns signed evidence bundles and certification receipts. That means AUX is doing verification, not decision-making. It does not appear to keep a long-running session state or act autonomously over a workflow. Control goes back to the calling system, which decides whether to trust the evidence and proceed, or stop and ask a human.

See AUX's full scorecard →

One tool worth testing

Kane CLI — Kane CLI is worth testing for teams that need browser checks to produce reviewable evidence from the terminal. Its new workflow lets users describe tests in plain English, run them against real Chrome, and receive pass/fail results with replayable evidence packs—an expansion toward agent-native QA. In a 30-minute pilot, run one critical-path browser test and compare Kane’s replay package and pass/fail output with your current browser-testing workflow.

Our pick: n8n

The use case: workflow automation.

Tops our independent /36 — strongest on action and reliability, ahead of Zapier. Pricing: Per-execution (not per-step).

Ranked by our 9-dimension Agenticness rubric — not by who pays us (nobody does). The full scorecard, true-cost pricing, and the reliability evidence behind every score are in the report.

Read the full report →