AI agent infrastructure
An agent that can click, run code, or pay needs a box to do it in. This hub is sandboxes, headless browsers, frameworks, and traces. E2B and similar give you a VM. Browser tools give you a real page. LangGraph and friends orchestrate steps. If you only need chat completions, you want Models, not this list.
View rankingsBrowse alternativesBrowse by tags
Sandbox vs local agent?
Use a sandbox when untrusted model output might run commands. Local is fine for your own repo on your own machine (Claude Code, Cursor).
Do I need observability?
Yes as soon as more than one person depends on the agent. Traces beat screenshots of a demo that worked once.
Framework or just scripts?
Scripts until the graph branches. Then LangGraph or the vendor's own agent SDK. Do not start with a platform because a thread said so.
All AI agent infrastructure tools (59)
Open any tool below for details, features, and pricing.
- A2A ProtocolOpen agent-to-agent protocol. Spec is $0.
- Agent ReachOpen-source capability layer that gives AI agents one-click access to the entire internet β Twitter, Reddit, YouTube, Bilibili, XiaoHongShu, and more β with cookies kept local and fully open, auditable code.
- agentmemoryPersistent, searchable memory for AI coding agents with 95.2% recall on LongMemEval, running locally on SQLite with MCP, hooks, and 54 tools for Claude Code,.
- AgentOpsAgent session replay and traces. Free $0. Pro is $40/month. Enterprise is custom. GitHub is about 4.4k stars, not 5k+ unspecified.
- AgentryAgentry is an Apache-2.0 AMTP gateway that runs AI-generated multi-agent workflows and stores organizational decision traces.
- AlloomiMeland Labs workspace that turns expert methods into a self-evolving digital employee. Early access v0.6.1. Free, Basic $15/mo, Pro $39/mo.
- browser-useOpen-source Python library that makes websites accessible to AI agents, letting them drive a real browser to click, type, and fill forms.
- BrowserbaseHosted browser infra for agents. Developer is $20/month. Startup is $99/month. Scale starts at $249/month after a custom form.
- BrowserSkillTencent's MIT-licensed CLI plus browser extension lets a coding agent borrow one of your already signed-in tabs, then hand it back when the task is done.
- CC Safety NetA PreToolUse hook that blocks destructive Git and file commands and secret access before AI coding agents run them, defeating reordered flags, shell wrappers, and one-liners.
- Cloudflare ComputerPreview MIT agent filesystem on Durable Objects. Not production. Pay Workers usage, not a $0 computer SKU.
- Amazon CloudWatch OmniAWS's AI-first observability experience: one workspace to trace, evaluate, and troubleshoot applications and agents across accounts, regions, and clouds.
- cmuxOpen-source Ghostty-based macOS terminal built for coding agents: vertical tabs, notification rings when agents need attention, split panes, an in-app browser, and a CLI/socket API.
- CrewAIPython multi-agent framework with crews, flows, and an Enterprise Cloud from $99/seat. AMP is now that cloud, not a separate product name.
- CumoraCross-platform team chat where AI agents are first-class teammates. Cloud pods or BYOA Claude Code/Codex. Invite-only preview is free.
- DaytonaDaytona is now a closed-source sandbox cloud. GitHub is archived. Starter $20, Plus $200, Scale $500, plus compute after included credits.
- DeepSeek HarnessMIT agent runtime. Pay model API only. Live Flash peak $0.44/$1.32 per 1M, not the old $0.14/$0.28 row.
- DeerFlowByteDance's open-source SuperAgent harness for long-horizon tasks. Orchestrates sub-agents, memory, sandboxes, and skills on LangGraph, with a browser-equipp.
- E2BFirecracker microVM sandboxes for AI agents. Hobby is $0 plus usage; Pro is $150/month. Open-source SDK on GitHub.
- EmdashYC W26 ADE. Desktop $0; Cloud is contact.
- EveVercel's open framework for building production AI agents. Like Next.js for agents: instructions and skills in Markdown, tools in TypeScript, durable by default.
- Explee AutoGTMAutoGTM by Explee: a 24/7 AI sales agent that finds high-intent prospects, sends personalized outreach, handles replies, and books demos with pay-as-you-go p.
- FastClawLightweight AI Agent runtime and Multi-Agent Framework written in Go. An Agent Factory that creates, manages, and runs AI agents with personalities (SOUL.md).
- Claude for Financial ServicesAnthropic's official plugin pack for finance: reference agents and skills for pitch decks, DCF/LBO models, earnings review, and GL reconciliation.
- Google ADKGoogle's agent SDK in Python, Java, Go, and TypeScript. Docs now live at google.github.io/adk-docs. A2A is a separate protocol repo.
- Grok BotSpaceXAI's always-on AI teammates that run on their own cloud computer, sign into your tools, and finish work end-to-end β available in beta for SuperGrok Heavy, Cursor Ultra, and Cursor Teams Premium subscribers.
- Grok BuildxAI's Rust CLI coding agent. The live GitHub repo is superagentic-ai/grok-cli. Auth is Grok.com or an xAI API key.
- Hark HandoffComputer use agent that navigates any website end-to-end on a dedicated virtual computer. Book flights, order food, shop, and message candidates on LinkedIn.
- HeliconeOpen-source LLM gateway and traces. Cloud Free is $0. Pro is $79/month. Team is $799/month. Extra logs start at $0.28 per 100k.
- HerdrOpen-source runtime that keeps your coding agents alive: real terminal sessions on a background server, so Claude Code, Codex, and opencode keep working when.
- Hermes AgentSelf-improving open-source AI agent by Nous Research with a built-in learning loop, multi-channel gateway, and model-agnostic runtime. MIT licensed.
- KitesurfCloudflare agent browser on Workers. GitHub 404. Not an open-source Chromium clone.
- LangfuseOpen-source LLM tracing and evals. Cloud Hobby is $0. Core is $29/seat. Pro is $199/seat. Plus usage after included units.
- LangGraphLangChain's low-level agent orchestration runtime: durable graphs, streaming, human-in-the-loop, and persistence. About 39,835 GitHub stars.
- LangSmithLangChain Cloud for traces, evals, and deploys. Plus is $39/seat. Startup is $1,400/month. LangSmith Engine is $2,000/month.
- LettaMemGPT's company and OSS agent memory runtime. GitHub letta-ai/letta is ~20.5k stars. Cloud is $20 and $200 per project plus usage.
- LuvusLuvus is Apache-2.0 Rust mission control that watches, resumes, and orchestrates coding agents in one TUI.
- MastraOpen-source TypeScript agent framework: agents, workflows, RAG, memory, MCP, and eval with a hosted platform. About 27,437 GitHub stars (2026-08-25).
- ModalPython serverless compute with a Sandbox API for untrusted agent code, scale-to-zero, and per-second CPU/GPU billing.
- MulticaOpen-source workspace that assigns issues to coding agents as teammates. Self-hostable, 26 CLIs, daemon runs on your machine.
- Microsoft MXCMicrosoft Execution Containers: an open, cross-platform sandboxed code-execution system for running untrusted model output, plugins, and agent tools.
- nanobotUltra-lightweight, self-hosted personal AI agent framework in Python with WebUI, tools, long-term memory, MCP integrations, multi-agent workflows, and scheduled automation.
- NVIDIA NOOANVIDIA's open-source agent framework that turns an agent into one Python class: methods are actions, docstrings are prompts, at 82.2% SWE-bench Verified.
- Open AgentsVercel Labs' MIT-licensed reference app for running background coding agents: a web UI, a durable workflow runtime, and isolated sandboxes you can fork.
- OpenAI Agents SDKOpenAI's official Python and TypeScript agent runtime: agents, handoffs, guardrails, sessions, and tracing.
- OpenSquillaToken-efficient microkernel AI agent with ML-based smart routing, persistent memory, a layered sandbox, built-in search, and local embeddings. Apache 2.0.
- Pizza BotPizza Bot is a local-first Apache-2.0 inbox for long-running AI agents, with Unread and Action queues on DeepAgents and LangGraph.
- PraxistPraxist is a Fair Source research runtime that runs parallel peers, durable evidence, and Codex or Claude Code skills on a measurable project.
- ReasonixDeepSeek-native AI coding agent for the terminal. A config- and plugin-driven harness in a single static Go binary, tuned around DeepSeek's prefix cache to k.
- ShipAnyAgent-native AI SaaS template family: auth, payments, subscriptions, credit billing, i18n, storage, and AI generation pre-wired so coding agents can ship AI.
- SoL-PiSoL-Pi is NVIDIA's MIT Pi extension that cuts agent tokens with four opt-in mechanisms and no patch to Pi itself.
- SteelOpen-source browser API for agents. Cloud Launch is $0 plus usage with a one-time $30 credit. Scale is $250/month plus usage.
- SupersetSource-available desktop app (Elastic License 2.0) to run an army of CLI coding agents in parallel β Claude Code, Codex, OpenCode, and more β each in its own.
- Microsoft ThinkingBoxOpen-source Microsoft sandbox and benchmark that verifies AI agents reliably complete stateful business tasks, exposing the discovery-reliability gap.
- TodosTask-driven workspace where a Chief agent splits your goal into todos, assigns each to agents by trade, and tracks progress β supporting 700+ models, including ChatGPT/Codex or GitHub Copilot sign-in without an API key.
- Trigger.devOpen-source durable execution platform for AI agents in TypeScript. Build long-running tasks and chat agents with no timeouts, checkpoint-based resume, retries, queues, and idempotency.
- UI-TARS-desktopByteDance's open-source multimodal AI agent stack for controlling computers and browsers, with Agent TARS CLI, remote operators, and the UI-TARS-1.5 model.
- Vast.aiGPU rental marketplace with interruptible and on-demand instances. Prices move by host. Confirm the live offer before you rent.
- Vibe SquadMulti-model coding harness: one coordinator routes 71 Markdown specialists across Codex, Claude, Gemini, Grok, and Kimi in git worktrees.