Tag
106 posts tagged #ai-agents
Coverage focused on agent frameworks, orchestration patterns, memory systems, evaluation loops, and practical AI-agent tooling for developers.
webctl flips the MCP browser automation model on its head: a CLI tool that gives you full control over what enters your AI context window, with benchmarks beating agent-browser on quality and cost.
A remote MCP server that lets AI agents search, publish, fork, and improve browser games on GameFork — TIC-80 cartridges, HTML games, and more.
Safari MCP drives your real Safari browser from any MCP-compatible AI coding agent. 97 tools, no Puppeteer or Playwright needed, ~60% less CPU than Chrome.
Open-source gateway that stores API credentials once and injects them transparently for AI agents. Never give agents your real keys.
Cordon is an open-source security gateway that sits between AI agents and MCP servers, enforcing policy-based access control with human-in-the-loop approvals.
Open-source observability for MCP servers. TypeScript and Python SDKs plus a CLI, with real-time tracing, per-tool latency breakdown, and alerting.
Klavis AI provides MCP integration platforms that let AI agents use 100+ prebuilt tools reliably at any scale, with self-hosting, Python and TypeScript SDKs, and REST API. Apache-2.0 licensed.
Smithery connects AI agents to 300+ MCP tools — weather, crypto, search, DNS and more — with auth and credentials handled for you.
Cloud-hosted AI agents that control a real browser to automate web workflows — prospecting, outreach, social media, and lead generation without code.
TopSend charges $20/month for unlimited email campaigns, transactional emails, and AI agent API access — no per-contact fees.
Browser Harness connects an LLM directly to your Chrome browser via CDP, letting the agent write its own helper code when tasks require custom logic.
A practical guide to Nanobot, the open-source standalone MCP host that lets you build dedicated AI agents from any MCP server with a YAML config.
Kybernis adds an idempotency and execution ledger layer to AI agents. Prevents duplicate refunds, repeated mutations, and inconsistent state. Framework-neutral.
Open-source platform for building multi-task AI agents with a no-code interface, PostgreSQL backend, and support for OpenAI, Anthropic, and Gemini models
l6e is an open-source MCP server that adds cost enforcement to Cursor, Claude Code, and Windsurf. Set a session budget, and the agent checkpoints and halts when spending gets too high. Free MIT tool.
agentmbox gives AI agents their own email inbox without sign-ups or API keys. Agents pay per request via x402 USDC on Solana or Base — $0.50 to create a mailbox, $0.005 to send.
An analytics workspace built for AI agents. Connect data sources, build dashboards with Claude/Cursor/ChatGPT, and share durable traceable reports instead of one-off chat threads.
Actionbook turns any website into an AI-agent-operable page via MCP connector and Chrome extension. No CLI install, handles logins and paywalls, 10x faster than traditional browser agents.
Browser Use is an open-source AI agent framework with 99K GitHub stars that controls real browsers via Python. This guide covers setup with Cursor and Claude Code, CLI usage, and when to pick open-source vs cloud.
AIMX is a self-hosted SMTP server with built-in MCP integration. Give AI agents their own email addresses with Markdown-based storage and DKIM trust.
YC-backed endpoint security for AI coding agents. Monitors Cursor and Claude Code at OS level with eBPF and ESF. Tracks file access, network, and processes independently.
Deploy CrewAI, LangGraph, and LangGraph.js agents with a single command. Real-time streaming, auto-scaling, and zero infrastructure to manage.
Build and deploy durable AI agent workflows in TypeScript. Open-source Zapier alternative with retries, queues, observability, and self-hosting.
Open-source TypeScript framework for building AI agents with memory, RAG, MCP, and multi-agent workflows. Includes VoltOps observability console.
Build, run, and export MCP tools and AI agents entirely in your browser. No install, no server, no backend. Powered by Pyodide WASM and DuckDB-WASM for free.
Awesome Architecture is a bilingual architecture knowledge base with 26 tutorials, 25 system maps, and 6 end-to-end cases for developers designing AI and distributed systems.
Google's Antigravity SDK packages typed tools, streaming responses, MCP integration, hooks, and triggers into one Python agent runtime for production-grade workflows.
OpenSquilla combines routing, planning, memory, and optional MCP support into a token-efficient local-first agent workflow with practical CLI setup and tradeoffs.
Browse 5,000+ MCP connectors for Claude, ChatGPT, and Cursor. Glama provides a live registry with uptime monitoring, direct connection, and a gateway for remote MCP servers.
Obot is an open-source MCP gateway that manages MCP servers, Skills, access policies, audit logs, and integrations for AI agents. Deploy via Docker, connect to any MCP client.
Spec27 from Safe Intelligence is a spec-first testing platform for AI agents. It generates adversarial tests from declarative specs and validates vendor agents without SDK access.
GitAgent turns any git repo into a portable AI agent definition. Framework-agnostic, version-controlled, with Claude Code, OpenAI, and CrewAI adapters.
YourGPT 2.0 unifies chatbot, voice, and helpdesk into one AI agent platform with native MCP marketplace support, multimodal inputs, and 20+ integrations.
Mercury Agent Skills collects reusable packs for Mercury Agent, OpenClaw, and Hermes so teams can add memory, workflows, and domain playbooks faster.
agentcookie syncs Chrome cookies, CLI bearer tokens, and API keys between Macs over Tailscale so agent machines stay authenticated without repeated logins.
WeSight wraps Codex, Claude Code, OpenClaw, and other local agent CLIs in one macOS workspace with model routing, runtime metrics, and skills.
Query multiple data sources with GraphQL-style syntax without a data warehouse. Nerve stitches SaaS APIs into one mega-API for AI agents and internal tools.
Superglue (YC W25) turns natural language into production API integrations. Build workflows connecting Stripe, Salesforce, legacy SOAP, and more with schema drift detection and MCP support.
SuperHQ runs Claude Code, Codex, and custom AI agents in isolated microVMs on macOS. Each agent gets its own sandbox with secure auth gateway and diff review.
ContextFort is an open-source Chrome extension that gives security teams visibility and granular controls over AI browser agents like Claude in Chrome.
Run AI agents in hardware-encrypted TEEs backed by AMD SEV-SNP. Private inference, vTPM-bound keys, and TEE-attested execution for OpenClaw workloads.
Open-source MCP testing platform: Playground, OAuth conformance, evals across models, and a Skills tab that teaches agents how to use your MCP tools.
Pando is a code firewall for AI agents that indexes your AST and gives Claude, Gemini, and Codex compiler-checked edits, atomic snapshots, and 100x token savings.
Eve is a managed OpenClaw agent harness that runs billing, support, releases, and revenue ops across 3,000+ integrations from iMessage, Slack, and email.
Stagewise is an open source agentic IDE (AGPLv3) that ships a coding agent with full tab console/debugger access, supports 25+ models via BYOK, and orchestrates agents across git workflows.
Entire ships Checkpoints that capture AI agent sessions as versioned Git metadata, backed by $60M from Felicis and led by ex-GitHub CEO Thomas Dohmke.
Open-source trust infrastructure that gives AI agents behavioral contracts, real-time integrity monitoring, and cryptographic trust ratings across multi-agent systems.
Rivet is an open-source desktop IDE from Ironclad for building, debugging, and deploying LLM prompt graphs. Ships with a TypeScript runtime library.
BracketMadness.ai is a March Madness bracket challenge designed for AI agents, not humans. Agent-first UX, plain-text API docs, and a real leaderboard ranking models.
CSS Studio is a browser-based visual editor that streams layout and style changes to Claude Code, Codex, and Cursor via MCP. From the Motion team.
Rtrvr.ai exposes your Chrome as a remote MCP server so any AI agent can drive it. Learn how the architecture works and how to set it up in minutes.
Construct Computer gives autonomous AI agents a persistent cloud desktop, an email address, and live browser access so they can run real work schedules, not just one-off API calls.
liblab MCP Generator turns any OpenAPI spec into a deployed, hosted Model Context Protocol server in 30 seconds, ready to plug into Claude, ChatGPT, or Cursor.
Ink is a deployment platform designed for AI agents, with MCP, Skills, and CLI integrations. Agents ship services in under a minute, billed per-minute while running.
JungleGym is an a16z open-source playground with three web-agent datasets and TreeVoyager DOM parser for benchmarking and training autonomous agents.
Cekura is a testing platform for voice and chat AI agents with simulated conversations, LLM-judge and code-based evaluators, production monitoring, and CI integration.
Airtop lets AI agents drive cloud browsers via Puppeteer, Playwright, or natural language. Handles bot detection, proxies, and captchas. Full setup guide, pricing breakdown, and alternatives.
Sonarly (YC W26) deduplicates noisy Sentry and Datadog alerts, runs root-cause analysis, and ships fix pull requests through Claude Code instead of paging on-call.
One (formerly Pica) is a Rust-based open-source agent infrastructure layer that gives any AI agent managed auth, audit logs, MCP access, and 70,000+ tools across 400+ apps via one CLI.
Lume is a free open-source CLI to create macOS and Linux VMs on Apple Silicon. Zero-touch setup, MCP integration, isolated sandboxes for AI agents. Native Virtualization Framework — no emulation.
Local-mode multi-agent matrix framework that runs supervised Claude Code or Codex agents — coordinated through JSON-RPC, per-agent worktree isolation, and async message bus.
Version control purpose-built for AI coding agents — blame which prompt wrote any line, inspect full change context, and track every agent turn as a content-addressed step.
Persistent memory layer for AI agents across sessions — 9-stage consolidation turns observations into facts, relationships, patterns, and wisdom over time.
AWS skills and MCP server for Claude Code, Codex, and Kiro — covers 300+ AWS services through CDK, CloudFormation, serverless, containers, and Bedrock agents.
Self-improving context layer for data agents — ingests databases, BI tools, and wiki content; builds semantic layer with automatic fan/chasm trap resolution.
Ingest documents into persistent navigable memory — tree-like hierarchy reconstruction, multi-modal parsing, and agentic RAG with evidence-based citations.
AnySearch MCP Server gives AI agents real-time web, finance, and academic search through MCP tools, with batch queries, URL extraction, setup steps, and practical use cases.
Animated desktop pets that react to Claude Code, OpenCode, and MCP agents — privacy-safe reactions and speech without exposing code or secrets.
AI agents pay per email request via USDC on Solana or Base and get their own inbox with IMAP, SMTP, and REST API. No signup, no API keys — just @agentmbox.com.
Hopper connects AI agents to your mainframe through Model Context Protocol, enabling natural language operations and autonomous workflows on COBOL/Java systems.
AI agents that autonomously monitor, root-cause, and remediate GPU infrastructure issues. Reduce compute costs, improve GPU utilization, and accelerate ML research. YC W26.
Runtime gives your whole team safe access to Claude Code, Codex, and other coding agents with snapshot environments, scoped secrets, and infrastructure-level.
AI agents that automate intake forms, chart prep, and referrals in healthcare clinics. HIPAA compliant with deterministic checks.
Sim Studio is a collaborative visual interface for building and deploying AI agent workflows. Design agents with a node-graph canvas, deploy as APIs, and.
YC W26 startup builds deployment infrastructure for sandboxed coding agents, research agents, and document processing tools that need filesystem access to work.
Mastra is an open-source TypeScript framework for building AI agents. Learn how it handles workflows, memory, RAG, MCP integration, and local debugging for production apps.
Klaus runs OpenClaw—the 375k-star open-source AI agent—on a dedicated EC2 instance with pre-configured OAuth, automatic hotfixes via ClawBert AI SRE, and zero.
Turn job descriptions into projects, projects into resumes, and resumes into interviews. AI-driven internship prep tool for CS students — from JD analysis to.
Spec-driven coding harness with self-improving context memory. 12 specialized agents, 32 skills — kills context rot and ships features instead of spaghetti.
CLI, SDK, and IDE plugins for adversarial AI agent workflows. Pit two AI agents against each other — one builds, one critiques — for higher-quality code.
Omnara is a web and mobile agentic IDE that lets you run Claude Code and Codex sessions from anywhere. Keep your coding agents running even when you're away.
Minicor connects AI agents like Claude Code and Codex to Windows virtual machines through an MCP server, enabling scalable desktop RPA with Python workflows.
Ardent gives coding agents production-like Postgres sandboxes instantly, so they can test SQL safely without risking your actual database.
Trace, debug, and evaluate AI agents in production with Lucidic's observability platform — one-line init, graph visualizations, and time-travel debugging.
Voygr provides real-time place intelligence for AI agents—business validation, freshness signals, and web context that Google Maps APIs cannot surface.
Canary reads your codebase, understands what a pull request changed, generates end-to-end tests for affected user flows, and runs them against preview apps.
Strata is an open-source MCP server that reveals AI tools step-by-step as an agent works, instead of overwhelming it with thousands of tools at once.
SourceTable Superagents lets you connect any spreadsheet to external databases, REST APIs, or MCP servers — turning Excel into a universal data hub powered by.
Sentrial monitors AI agents in production, detecting loops, hallucinations, and tool misuse the moment they happen. Here's how it works and how to set it up.
Mozilla-backed API that abstracts complex web browsing for AI agents—handles proxies, hydration, DOM parsing, and returns clean structured data instead of raw.
Notte records your browser interactions once and compiles them into deterministic automation code — no LLM at runtime, no fragile selectors.
Dedalus Labs spins up persistent full Linux VMs for AI agents in under 250ms. No Dockerfiles, no YAML configs — just SSH, GPU, and a Python SDK.
Propolis runs swarms of AI browser agents that explore your web app, find bugs, and generate Playwright e2e tests for your CI pipeline. Here's how it works.
Hyperbrowser spins up hundreds of headless browser sessions in secure isolated environments with sub-second launch times, captcha solving, and residential.
Dedalus Labs connects any LLM to any MCP tool via a single OpenAI-compatible API endpoint. Upload MCP servers to the cloud, skip the Dockerfiles and YAML.
Twill.ai is a YC S25 startup that lets you hand off coding tasks to cloud AI agents and get back finished pull requests on GitHub—no local setup required.
Voker is an agent analytics platform that gives engineering teams structured visibility into how their AI agents behave, perform, and cost — without digging.
Spine Swarm runs parallel AI agents across 300+ models simultaneously, enabling research and client-ready deliverable creation through a visual canvas.
Kita uses vision-language models to automate document-based credit review for lenders in emerging markets, parsing 50+ document types from PDFs to photos.
Smooth CLI replaces low-level click/type/scroll with natural language commands. 20x faster, 5x cheaper, and works with Claude Code, OpenClaw, and any AI agent.
Cloud browsers built for AI agents that need to navigate, scrape, and interact with the web at scale. Handles proxies, stealth detection, and concurrent.
YC W26-backed AI agent observability platform. Trace sessions, detect silent regressions, and A/B test prompts in production before failures reach users.
Superset runs Claude Code, Codex, Cursor, and other AI coding agents simultaneously in parallel workspaces. Orchestrate agents, automated workflows, and code.
Define multi-agent AI workflows in YAML and run them locally with one command. AgentMesh brings Docker Compose patterns to AI agent orchestration.
Evaluate Photo-agents for image-agent workflows, including license keys, Python isolation, sample-image testing, metadata checks, and batch safety.
Explore awesome-agentic-ai-zh as a Chinese agentic AI learning roadmap, with setup notes, track selection, study workflow, and evaluation guidance.