Lindy

General-purpose autonomous agent that completes multi-step web tasks.

Open-source self-improving agent from Nous Research with persistent memory, skills and messaging gateways.

Open-source browser agent library (120k+ GitHub stars) plus stealth cloud browsers at $0.02 per browser-hour.

Hosted persistent sandboxes for agent harnesses like Hermes, OpenClaw, Claude Code and Codex; pay per minute of uptime, no subscription.

Multiplayer AI agent builder: anyone on a team builds agents with any model and integration while IT controls access; Pro from $37 a month.

Cloud browsers for computer-use agents, with managed sign-in and an Action Index that replays proven web workflows as code. Free tier available.

Open source (MIT) desktop AI team for macOS and Windows: a Commander agent splits a goal into to-dos and coordinates specialist agents plus local Claude Code and Codex. Free with 500 credits.

Coordination layer for teams of AI agents. Its new desktop app shows what each coding agent is working on and where it is stuck, and lets agents hand work to each other across tools and machines.

Approval layer for AI agents. Intercepts tool calls from Claude Code, Codex, MCP and agent frameworks, approves routine ones by rule and sends risky ones to a person. Free hosted tier; self-hostable core under the Elastic License 2.0.

A team of AI agents for non-technical people. Each agent gets its own computer and the apps you choose, shows each step and asks before anything is sent, bought or deleted. Starter $25 a month.

Open-source MCP server with 31 web tools (scrape, crawl, stealth browse, multi-source research) for Claude Code, Cursor and other agents. Clean markdown or JSON; 1,000 free credits to start, then pay per use.

Tencent Hunyuan's local-first memory app for agents. Imports and syncs chat, work and coding memory across tools like Cursor, Claude Code and Codex so preferences and decisions carry over. Part of the open-source Hy-Memory stack.

Tool and auth layer for agents: gives Claude, Codex, Cursor and other MCP clients access to 1,500+ apps through SDKs, a CLI and an MCP gateway. New Instant wallet pays for tools like search, voice and scraping per call with no markup.

Capability layer that plugs Codex, Claude Code, Cursor, ChatGPT and other MCP clients into apps, 27,000+ actions and 400+ skills, with credentials kept in a vault and an audit log of what ran and what it cost. Pay as you go from $0 a month; new, so the rank can move.

Free, open-source (MIT) command center that runs Claude Code, Codex, Cursor, OpenCode and other coding agents side by side across your machines, each on its own git worktree, with end-to-end encrypted links. Mac and Linux; launched Oct 15, 2026.

Claude Command Center: a free, local dashboard that puts every Claude Code, Codex, Cursor, OpenCode and other coding-agent session on one board, even ones you launched by hand, and flags the ones waiting on you. Tracks usage limits and can fall back to free models. Source-available (FSL-1.1-MIT); Claude Code is first-class, other engines have documented gaps.

Gives AI agents a real phone to drive mobile apps through MCP or an API: cloud Android phones billed monthly or per minute, plus a robot-tapped iPhone sold by call. Niche, but it reaches apps that have no API.

Lets personal AI agents use your saved logins without seeing them: the agent asks, you approve on your own device, and 1Password fills the sign-in form. No extra cost on 1Password Individual and Family plans; agent integrations are still rolling out.

Cloud browsers for AI agents on Moli, a browser engine Lexmount built for agents, with a Chromium option plus Web Search and Web Fetch APIs. Free plan includes 10 browser hours a month. Newer and less proven than Anchor Browser.

AI customer-support agents for email, chat, voice and WhatsApp that can act in your helpdesk or store, such as processing returns and order updates, and hand off to a human with context. Free Starter plan with 300 credits a month; Pro from $25 a month. A niche pick here: only worth it if you run a support team.

An AI to-do list app where you hand a task to Claude, ChatGPT or any MCP agent. It shows its plan, asks when unsure and brings back the work with proof, and nothing is marked done without your OK. Free plan with 100 agent calls a month; Pro is $5 a month. Launched in October 2026, so it is new and has little independent track record.

Pay-per-task AI agent that works across Gmail, Drive, Slack, Notion and Calendar and shows an itemized receipt for each task. No subscription; top-ups from $5 never expire.

Local guard for coding agents like Claude Code, Codex and Cursor that can allow, ask about or block risky shell commands and secret access before they run. Free forever on one machine; cloud sync is paid.

Shared operations board for teams of AI agents: tasks, messages, hand-offs, human approvals and an audit trail, over REST or MCP. Free during public beta; paid plans not open yet.

Shared company context for the AI agents you already use: connect inboxes, docs and apps once, and Claude, ChatGPT or Cursor reach it over MCP with cited answers and reusable skills. Free plan with 200 credits a month; paid from $20 a month. New, from Devcore.

20 free, open-source skill files by Poptin that teach an AI agent such as Claude, Codex or Cursor to plan email campaigns, popups, automations and segments by industry. Instructions only, it sends nothing; a brand-new repo, so expect changes.

Home for AI artifacts: agents publish HTML, Markdown, YAML or JSON over MCP or API to permanent sandboxed share links with comments, versions and access controls. Free plan with unlimited artifacts; Pro normally $6 a month (first 500 accounts get Pro free for 12 months).

Persistent cloud Chrome for AI agents with isolated profiles, CDP links, pause, resume and fork. Free plan with 1 browser-hour a month; Basic $20 and Pro $99 a month. Newer and smaller than Anchor Browser.

Marketing plugin for Claude and ChatGPT: outbound, SEO and ads tools over MCP, with a no-card trial.

Shared workspace where Claude, ChatGPT, Codex, Grok and other MCP agents keep files, tasks and pinned decisions, and anything they want to send or post waits for your approval. Free for one person; Pro is $20 per member a month. New and similar to Pragor and AgentOS, so the rank can move.

Email API that gives each AI agent its own inbox: one request returns an API key and an address, inbound mail arrives as a signed webhook, every message is full-text searchable, and there's a hosted MCP server. Free with 5 inboxes and 5,000 emails a month; Starter $12 and Pro $99 a month. New and small.

Free coach and security-scanned library for AI agents: skills, MCP servers and prompts scanned at a pinned commit, matched to your agent's one goal, with daily check-ins that say what to keep or drop. Connect over MCP or REST. New, and from the team behind deed.page and Cooper Email, whose own tools sit in the library.