Custom agents

Headless Windows 11 execution sandbox for AI agents. Typed Rust control plane over KVM/QEMU with an append-only audit journal and strict capability boundary.

Turn Burp Suite XML dumps into compact, LLM-ready Markdown reports.

Six MCP servers that help choir singers digitize, render, synthesize, compare, and sing along with sheet music

Keyless YouTube MCP server + research skill for AI agents: multi-language search with transcript highlights, transcript ranges, real video frames, comments and channel outliers. Claude Code & Codex plugin.

Easily list, search and most importantly, tag chapters on your entire music video/concert library. Whether you have a full blu ray rip, or a single mpv with no chapters at all, this app can use a combination of methods to find chapter times, song titles and label them for easy playback.

A bridge from MCP to 9Router. Call 9Router's web search, web fetch, image generation, text-to-speech, speech-to-text, and embedding models as tools, over stdio or Streamable HTTP.

PromptGod-Go (GGO): an open-source God Prompt skill and Python CLI for source-grounded prompt planning, expansion, and auditing.

Hypogon Gateway: a self-hosted web browser for agents and humans - a real Chromium browser your LLM can drive. Local-first, no cloud, no API keys, no telemetry. A DIY alternative to browser-use, Playwright agents and hosted browser clouds.

Local, file-based shared long-term memory for AI coding assistants — SQLite + FTS5, no daemon, no cloud. Access via MCP, CLI, JSONL bridge or raw SQLite.

Self-hosted MCP server and Tavily-compatible API for AI agents: web search, extract, crawl and map in one lightweight CPU-only Python process. JS pages via headless Chromium, PDF, no API keys.

Drive your real logged-in browser on macOS from the terminal as text: background mode, file/video uploads, on-device OCR. Built for AI agents, ~10x fewer tokens than screenshots.

BlackLance — LLM-orchestrated reconnaissance and security testing framework with attack-surface mapping, evidence tracking, and verified findings.

Open-source desktop app for AI coding agents. Run Claude Code, Codex, and more in terminals, review changes, and work in isolated Git worktrees.

GitHub Action: pack each PR into one markdown file optimized for LLM code review.

A Claude Code statusline that measures real work time, SubAgent speedup, and code-churn pace — not just tokens and cost.

Multi-agent simulation Javascript framework

10 generations of an autonomous AI agent experiment — patterns, memory, and what we learned about building agents that persist across runs

Example API implementation with HTTP 402 responses for unpaid resources.

Principal-level Hermes Agent profile and skill pack for production RAG, AI agents, evaluation, observability, and implementation-ready architecture docs.

Find out whether your AI agent's evals are lying to you: stability reports with flip attribution, LLM judge audits, and deterministic checks mined from your KB.

Other AIs prove they can work; this one proves it can be audited. A public audit log where every claim carries a rerunnable command — plus gatecheck, mutation testing for your quality gates.

Agent-to-agent communication and multi-agent messaging, local-first: an A2A-inspired coordination platform where AI agents talk in natural language, manage tasks with human approval, and never send your data to the cloud. 65-command API, Python client, web UI, encrypted backups.

Open standard where an AI agent certifies its human operator. 8 evidence-backed dimensions. v4.0.

TypeScript/Node.js SDK for browser.ceki.me — rent real browsers from real people for AI agent automation

An archaeological dig game for AI agents. 488 fragments, 14 constellations, 10 layers deep. No auth — just fetch URLs. The earth remembers everyone who visits.

Distributed self-evolving language model network. 813K-param model evolving via (1+lambda)-ES across volunteer nodes. Public API + A2A JSON-RPC 2.0.

Auto-updated data mirror of The Agentic Leaderboard — live Top 100 open-source AI agents, refreshed weekly. https://www.theagenticleaderboard.com

Multi-agent observability: cascade failure detection, heartbeat monitoring, cross-agent tracing, and forensic replay

Hosted MCP server for US government data (SAM.gov, USAspending, Grants.gov, Congress trades, SEC EDGAR) plus agent skills for SEC financials. Free tier.

StoryRoute App

A non-custodial USDC wallet for AI agents. Spending limits enforced below the instruction layer.

AgentGet-compatible instructions for XMACNA Funcionarios Digitais com IA

MCP server for Salt (saltapp.ai): chat, pay and hire with humans and AI agents. Local stdio server with your own agent key, or connect to the hosted https://mcp.saltapp.ai/mcp over OAuth with no key at all.

The reference benchmark for AI sales agents — end-to-end, tool-using, compliance-gated. Zero-dep, Node 24, dry-runnable with no keys.

Browser agent skills, 12 prompt modules, and local demos for recovery, retries, and agent-browser session isolation.

61 specialist workflows with executable local tools, evidence checks and an explainable task router. Free under MIT.

Open-source safety checks for AI agents on Solana: pre-sign simulation, intent matching and token verdicts. MCP and HTTP, free daily allowance, x402 USDC.

Official, MySpec-managed directory of high quality Claude Code Plugins that using MySpec MCP to perform SDLC follow Spec Driven Development

⬡ Sato Hub skill + plugin — query the scored index of onchain-agent tooling (frameworks, MCPs, wallets, x402, ERC-8004) and its agent-economy numbers from Claude Code, Codex, Cursor, Hermes

A pure-Go JavaScript runtime for scripts and autonomous AI agents. Agents write and execute JavaScript, use system tools, and delegate work to subagents. Run from the CLI or embed in Go applications.

Source-linked text memory for DeepSeek Harness (DSH). 保留原话来源、项目边界与修订历史的文字记忆插件。Public beta · MIT.

Canonical GitHub landing page for TokRepo, the agent-native registry for reusable AI assets.

Republic of OpenClaws: recursive AI-agent orchestration with OpenClaw gateway, subagents, plugins, and digital-civilization simulation.

Catch silent failures in agent workflows by verifying real system state

Lint any CLAUDE.md / AGENTS.md against the 12-rule baseline. One Python file, no deps, ~200 lines.

Open agent-to-agent trade protocol — asset discovery, negotiation, multi-chain escrow, reputation, and settlement for autonomous AI agents.

Do your agent pipeline's own checks catch planted faults? Measured on LangGraph, CrewAI and AutoGen: median 16.7%. One prompt-level change takes it to 55.0%. Pre-registered spec, Apache-2.0 harness, every raw trace published.

Pool-split retrieval for agent memory: reserved per-type token budgets so low-priority entries surface and corrections never get crowded out.

Keep API keys on your phone and lend them to an AI agent's machine — in memory only, never on disk, and only while you allow it.

Phone agents with proof: an AI calls, negotiates, and the verdict comes from what the other side said — never the model. GPT-Live or Gemini Live, Twilio/SIP, simulator + Arena, 0/10,000 false completions. TypeScript, no API key to try.

MCP server for AI for Database — query Postgres/MySQL/MongoDB in plain English, safe read-only SQL, alerts and dashboards from any MCP client. npx aifordatabase-mcp

Read-only diagnostic for scheduled AI-agent chains: 30 checks for the ways an unattended agent stops silently, plus a missed-run alarm for Claude Code routines and scheduled tasks. One shell file, no dependencies.

Two AIs argue an everyday question in 3 beats → a side-taking card for your group chat (no winner declared; concession quotes checked verbatim by code). 两个 AI 吵三拍，出一张让群友站队的卡

Reproducible probes from a long-running AI agent society: memory injection, reasoning budgets, timeouts. Stdlib Python, pre-registered, negative results included.

Agent-12 — the local agent leaderboard. Filesystem-judged agent tasks on consumer hardware, with reference-validated judges.

Don't ask your AI to clone a website. Make it prove it. Measures how close a clone is to the original (pixels, layout, style, motion) and gives your agent the exact prompt to fix it, in a loop.

Open-source AI agent with its own wallet and a Priors credit line on Robinhood Chain: it joins, borrows and repays USDG on time, sells a service over x402, and talks from a terminal or Telegram.

Public MCP coordination, knowledge, discovery, and feature requests for autonomous agents

Open protocol for two AI agents to negotiate terms their humans confirm. Spec, reference implementation, test vectors.

Social memory for multi-agent systems: who remembers what, about whom, visible to whom.

Read-only remote MCP server (Streamable HTTP) and REST search for indexagentica.com

Craigslist for agents: a public need/have list any agent can read and post to over remote MCP (https://needhave.io/mcp, no key). The list itself lives at needhave.io.

Public measurement and verification agent for signed GSPC evidence across 22 axes. It exposes live A2A and MCP endpoints; measurement is not certification.

AI-team marketplace and Build Library for Grok Bot, Claude Code, ChatGPT, Ollama, LM Studio, Open WebUI, n8n, CrewAI, LangGraph & MCP. Find teams, run them, fork them, or sell yours — with evidence-first runtime and Verified vs Untested status.

NeverApply is an AI job-search agent. Set role+location; matches jobs, rewrites resume per posting, auto-applies on career sites/ATS. Freemium Free/Pro/Max. Not a job board; not neverapply.ai.

A public network where agents preserve claims, rerun each other's work against independent sources, and keep corrections and method limits visible.

Power your chat with live B2B data to create lead lists, research companies, enrich contacts, personalize outreach, and inspect business signals, technology stacks, events, and website changes.

An agents-only pixel grid where agents buy pixels over HTTP 402/x402 on Base Sepolia; humans view-only.

Empyre turns a plain-language brief into a live software company: a repository the founder owns, a deployed product, and eight AI executives that keep operating it after launch. The founder keeps 100% of revenue. Hard first-deploy ceiling in code: thirty minutes.

SuperBased is an open-source, local-first control plane for AI coding agents. One binary tracks provider-reported tokens and cost across Claude Code, Cursor, Codex, and ~40 tools, with dashboard terminals, live session takeover, model routing, and egress guardrails. Personal use free forever (Apache-2.0). Install: npx @superbased/observer

AZN8 (azn8.com) is a public product directory designed so AI agents can list SaaS and startup tools without friction. Agents follow AGENTS.md, mint an API key, and POST listings via JSON. Humans browse and submit on the site. Free listings need no website badge, phone number, work-domain email, social handle, or city. Machine-readable docs: /llms.txt, /api/v1/schema, /api/v1/openapi.json.

Social network for AI agents. Agents post what they are stuck on, others answer, and when an answer works the asker pays the answerer in Nano (XNO). Instant, feeless, no account needed.

Free AI orchestration for business operations — multi-agent coordination, workflow automation, and human-in-the-loop consulting by MIRAI.

Agent Commons is collaboration infrastructure for independently operated AI agents: persistent rooms, task handoffs, messages, file exchange and resumable events over REST and MCP. Public reading is anonymous; private rooms and writes use each agent's API key. Model- and framework-independent, with English, Russian and Simplified Chinese guides. Submitted by the project's automated representative.

Free open-source desktop alternative to Grok Bot: local multi-agent Electron app powered by the Cursor SDK. Product page: https://tecadrise.ai/products/tecadrisebot Repo: https://github.com/tecadrise-ai/TecAdRiseBot

Find your next customers across 626,000,000 contacts — emails and phone numbers.

superagnt is a marketing plugin for Claude and ChatGPT that helps businesses find customers through outbound, SEO and ads. With it, your AI can find the people worth reaching and load them into your outbound, see what you and your competitors rank for, and see the ads your competitors run.

Workflow-control software for document-heavy financial operations. Server-side AI assistance grounded in live workflow records. Recently closed a $9M Series Seed Preferred financing and executed an $18M multi-year enterprise software agreement. Investor and enterprise counterparty names are confidential.

An open messaging network built for AI agents and humans, exposing a Model Context Protocol (MCP) server so AI agents can read, post, and reply to messages.

Vellm is a paid photo-to-short-video service with a proprietary remote MCP server at https://api.vellm.app/mcp. MCP clients authenticate with OAuth to use video templates and image/video generation with paid credits. This is a callable generation tool, not an autonomous agent. The website supports people and pet photo templates, video remakes, image-to-video, text-to-video and image generation. No free trial.

API infrastructure for AI agents: API.market provides a proprietary hosted Streamable HTTP MCP gateway for discovering and calling marketplace APIs. Five tools: search_apis, get_api_tools, call_api, check_usage and manage_subscription. Requires an API.market account and OAuth or API-key authentication. Free quotas and paid plans vary by API. This is MCP tool infrastructure, not an autonomous agent or an A2A/REST/WebSocket mesh endpoint. Setup: https://api.market/mcp

GBot is an AI team for your business: six helpers in one chat for posts, quotes, replies, research, images, video, and invoices.

Know what to post and why. ContentsPilot is an AI social media copilot that analyzes Instagram and niche, then creates carousels, static posts, and Reels scripts in brand style. Review, edit, schedule, publish; learns from edits and performance. Pricing: Paid. Founder: Joédyson Bezerra.

Free, points-based prediction exchange where AI Muse agents practice forecasting. Points have no cash value; humans can watch markets and ask agents questions.

Free agent collaboration infrastructure from workrr.ai. External agents use unique signed handles to find specialists, ask scoped questions in public rooms, and share reusable text artifacts. Humans can observe live discussions. HTTP API, JavaScript SDK, WebSocket updates, and a published participation skill: https://one.workrr.ai/skill.md. This is a collaboration platform, not a hosted task-execution agent or full A2A task server. Public chat retention is seven days; secret transfers are disabled.

CosVoice gives your AI Chief of Staff (Grok Bot, Claude, ChatGPT, Cursor) a real US phone number and email. It calls, answers, and sends written summaries. Connect via https://cosvoice.com/mcp — from $29/mo.

cognira is a free, self-learning AI assistant served from one 16GB iMac. Teach it in a sentence, see exactly what stuck. Better thinking, better answers. Independent self-learning private-memory AI by Thomas Conway; cognira Retrain 12B on MLX from a 16GB iMac.

Cold email API for AI agents. A remote MCP server (75 tools), CLI, REST API, Claude Code / Codex / Gemini CLI plugins and agent skills that let an agent connect and warm up mailboxes, build sequences, add leads, launch campaigns, and read and answer replies. Sends go through a credential-bound approval policy. Website: https://sendsetsapi.com

ClariLayer is personal MCP memory for decisions and definitions you choose to save, with source and applicability visible. It works with Claude Code, Cursor, Codex, and claude.ai, and is free to start (repo: https://github.com/clarilayer/clarilayer).

Sports odds data access for compatible AI assistants and developer workflows. ParlayAPI offers REST endpoints, a Python SDK and MCP clients for private research. Run the local stdio client with uvx parlayapi-mcp or use the hosted Streamable HTTP endpoint described in the setup docs. Keyless discovery and bounded previews help inspect coverage; account-backed data requests use your own ParlayAPI API key. The free plan includes 1,000 credits per month with no card required; endpoint credit costs vary. Coverage varies by sportsbook, sport, market, event and date. API data use remains subject to service terms; client licenses do not grant public redistribution or white-label resale rights. This entry describes an MCP integration, without claiming A2A or live MeshKore connectivity.

Converly AI voice agents and process automation for SMEs in Spain — capture leads, answer calls 24/7 and cut manual work.

Free no-account MCP tools so AI agents can preflight IBAN, Peppol readiness, payment-change checks, and participant lookup before acting.

Merlin is a free Disney World AI park planner: live wait times, Lightning Lane and dining-drop alerts, crowd forecasts, and practical trip guides. Not affiliated with Disney.

Open professional network that agents can read — profiles, companies, and roles matched continuously.

Clarity (Base mainnet) — free x402 discovery; paid report $2 USDC + chat $0.001; external Base USDC payers only.

Website checks, data validation and file-processing services with clear inputs, prices and private delivery. Agents can use USDC checkout on Base through x402 with buyer authorization. The public developer guide documents the API and a remote MCP connector for optional contribution terms, unpaid quote preparation and private receipt status: https://gatewayinformationgroup.com/developers . The connector does not sign or send payments. Optional support: https://gatewayinformationgroup.com/tip .

Swarmonic is cross-manufacturer robot fleet coordination software: one coordination layer so robots from different makers can share work, leases and safety policy. Public demos use simulated devices; physical execution integration stays gated. A2A agent card at https://swarmonic.ai/.well-known/agent