Current public demandBuild packet and draft write-up
MCP release sprint
Open-source maintainers and coding-agent builders are shipping against active MCP release motion right now. Give them a direct build path instead of a generic launch page.
Discover new launches, source-linked products, quickstarts, and adoption programs. Submit your own product and turn community attention into proof.
What builders are shipping
Each track below picks out a launch developers are already adopting, links the primary sources, and gives you a starting point to build against it and write up what you find.
Open-source maintainers and coding-agent builders are shipping against active MCP release motion right now. Give them a direct build path instead of a generic launch page.
63 organic products match the current filter set.
Labeled partner inventory. Organic radar sections below stay ranked by evidence and activity.
Fresh product additions and updates from the current day.
Wednesday, Sep 23
Recent tools and launch updates that are still fresh enough to check.
The newest approved tools in the graph.
HKUDS/nanobot · Open source
Ultra-lightweight, open-source, self-hosted personal AI agent framework in Python with WebUI, tools, memory, MCP, multi-agent workflows, automation, and chat apps
Vendors can claim a product, run public programs, track docs/demo clicks, and see which developer activity turns into account intelligence.
OpenAI and AI Gateway now support persistent Responses API WebSocket connections with previous_response_id continuation. Authenticate first, then keep that exact implementation query intact through stack search, save, Build Packet, and public proof.
The new @ai-sdk/harness-acp meta-adapter lets builders run ACP-compatible coding agents through one HarnessAgent interface. Authenticate first, preserve the exact runtime, MCP, session, and sandbox query in stack search, then save the recommendation into a Build Packet before publishing run proof.
Exa web search now works as an AI Gateway tool with any model and without a separate Exa API key. Authenticate first, preserve the exact grounding, citations, fallback, and trace query in stack search, then save the recommendation into a Build Packet before publishing research-quality proof.
Cloudflare Agents SDK now traces each agent turn across model calls, tool runs, approvals, token usage, and Workers runtime operations. Authenticate first, preserve that exact observability query in stack search, then save the recommendation into a Build Packet before publishing trace proof.
Current radar slice
Carry the launches you are looking at straight into a build draft or a proof page, instead of starting again from an empty workflow.
Saved by builders
proof state
2 saves · 48,337 stars
radar signal
Aidress-ai/Aidress · Open source
Open coordination and trust layer for autonomous agents with capability discovery, verification, matching, transactions, an SDK, API, and MCP server.
Source-linked
proof state
8 stars
radar signal
dinglebear-ai/soma · Open source
Production-oriented Rust RMCP runtime for shipping MCP servers with stdio and HTTP transports, REST, auth, observability, plugins, and generated contracts.
Source-linked
proof state
1 stars
radar signal
openwong2kim/wmux · Open source
Open-source workspace multiplexer for running and coordinating coding-agent fleets across persistent terminals, worktrees, browsers, and MCP tools.
Source-linked
proof state
388 stars
radar signal
every-app/open-seo · Open source
Open-source SEO and AI-visibility platform with self-hosting, pay-as-you-go data, MCP access, and reusable agent skills.
Source-linked
proof state
19,337 stars
radar signal
mcbuddy/testivai-oss · Open source
Local-first visual regression testing for Playwright and WebdriverIO with DOM-aware noise detection, machine-readable reports, and an MCP server for coding agents.
Source-linked
proof state
4 stars
radar signal
andrewyng/openworker · Open source
Open-source local-first desktop AI coworker with approval-gated actions, model choice, schedules, and 25+ work-app and MCP integrations.
Source-linked
proof state
18,031 stars
radar signal
risa-labs-inc/BossConsole · Open source
Open-source native operator console for running coding agents with browser, terminal, editor, secrets, RBAC, and 100+ MCP tools.
Source-linked
proof state
763 stars
radar signal
Ranked by saves, projects, adoption events, reviews, and public source signals.
apify/apify-mcp-server · Open source
Official Apify MCP server that lets AI agents discover and run Apify Actors as tools for web automation, scraping, and data extraction workflows.
Usage proof
proof state
1 project · 1 save
radar signal
ChromeDevTools/chrome-devtools-mcp · Open source
Official Chrome DevTools MCP server for browser inspection, debugging, automation, and performance analysis from coding agents.
Saved by builders
proof state
2 saves · 52,335 stars
radar signal
xai-org/grok-build · Open source
SpaceXAI's terminal coding agent with an interactive TUI, headless CI mode, MCP servers, skills, plugins, hooks, and sandboxing.
Source-linked
proof state
26,827 stars
radar signal
googleapis/mcp-toolbox · Open source
Open-source MCP server from Google that connects AI agents and IDEs to enterprise databases through governed tools.
Source-linked
proof state
16,467 stars
radar signal
microsoft/agent-framework · Open source
Open Microsoft framework for building, orchestrating, and deploying production AI agents and multi-agent workflows across Python and .NET with MCP/A2A support.
Source-linked
proof state
13,553 stars
radar signal
modelcontextprotocol/inspector · Open source
Official visual testing and debugging tool for Model Context Protocol servers.
Source-linked
proof state
10,930 stars
radar signal
tadata-org/fastapi_mcp · Open source
MCP adapter that exposes FastAPI endpoints as authenticated Model Context Protocol tools.
Source-linked
proof state
12,008 stars
radar signal
Launch intelligence
tokens& connects launch discovery, docs starts, project proof, and retained usage so enterprise teams know what to do next.
Free company workspaces require a work email. There is no pricing or self-checkout; paid enterprise access is provisioned manually after a sales call and agreement.