Spec Driven Development
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Description from the maker (original English)
Makes any AI agent globally reachable and discoverable—no cloud, domain, or port forwarding needed.
These three are never merged into one score: opinions, measurements and popularity are different things.
We have not written a plain-language summary for this one yet. Below are the facts we read from the registry.
Pick the AI app you use, then copy the settings.
You need
Which app do you use?
Pick your app above. We remember it, so next time the button says “Install in …”.
Open Claude Desktop → Settings → Developer → Edit Config, and paste this into claude_desktop_config.json
{
"mcpServers": {
"a2ald": {
"command": "npx",
"args": [
"-y",
"a2ald"
]
}
}
}Run this in your terminal
claude mcp add a2ald -- npx -y a2aldPaste into .cursor/mcp.json in your project, or ~/.cursor/mcp.json
{
"mcpServers": {
"a2ald": {
"command": "npx",
"args": [
"-y",
"a2ald"
]
}
}
}Your browser asks to open the app, then the app shows the settings and asks you to confirm. Key values are left empty for you to fill in.
Paste into .vscode/mcp.json in your project
{
"servers": {
"a2ald": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"a2ald"
]
}
}
}Your browser asks to open the app, then the app shows the settings and asks you to confirm. Key values are left empty for you to fill in.
This one runs on your computer, and ChatGPT can only connect to servers on the internet. Pick another app.
Your choice stays on this device in a cookie; no account needed.
These settings are generated from the maker’s registry entry. If the README says otherwise, follow the README.
Runs on your computer; the tool list appears in your AI app after you install it.
We check information anyone can see; we have not read the code. Checks that do not apply are left out of the count. · Checked Nov 6, 2026
The registry name was verified with the GitHub account a2al, which owns the repo
We have not read this repo from GitHub yet
We have not read this repo from GitHub yet
This one runs on your computer; there is no internet endpoint
Runs on your computer and uses local settings instead
Runs on your computer; we can only read tool lists from internet endpoints
Compared with every active server in this category in the registry. Stars count only when the publisher owns the repo.
It passed 1 of 1 applicable basic checks. We have not read the code. Before adding a key or token, see what access the server asks for and use a key with only the permissions you need.
The registry does not list a required key, though some servers ask you to sign in on first use.
It runs on your computer. Your AI app starts it when needed, and data does not pass through the maker’s servers (unless it calls an online service itself).
The GitHub repo last changed –, and the latest version (0.2.7) reached the registry 5 months ago.
Apps that support MCP, such as Claude Desktop, Claude Code, Cursor and VS Code. You also need Node.js on your computer.
In the same category: Spec Driven Development, Remote Desktop Commander, Apify MCP Server
Servers in the same category from other makers, sorted by stars.
Reading a card:★ GitHub starsiThree pillars: reviews · checks · usageiUpdated = latest version published to the registryi
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Remote Desktop Commander: use files, terminal and processes on your own computers from any AI app.
Extract data from any website with thousands of scrapers, crawlers, and automations on Apify Store ⚡
Copied from the maker’s repo as written. PopNx MCP has not reviewed or endorsed it. HTML, images and scripts are removed. Links lead off this site.
*Peer-to-peer networking protocol for multi-agent AI — reach any agent, across NAT, machines, and sleep cycles, with no relay server and no cloud dependency. Open source, MPL-2.0.*
Official sites: a2al.org · Tangled Network · tngld.net
Always reachable. Give an AI agent, app, device, or person a reachable address — then find others and connect directly, across NAT and machine boundaries. No domain. No cloud account. No central registry.
Your key, your address. That address is an AID: it comes from a key you hold. Nobody issues it, revokes it, or reassigns it. You hand it over once. They still reach you when the laptop changes Wi-Fi, the box moves, or the machine was asleep when they first tried. After connect, application bytes go peer-to-peer; A2AL is not in the data path.
The missing layer. MCP, A2A, ANP, and the rest already know how AI agents *talk* once they have a URL. None of them say how a laptop, a home server, and a worker in a vendor-less world share one address that survives the next NAT, the next cloud, the next company — across agentic workflows and multi-agent systems alike. That layer is missing. A2AL is that layer, and it has to live in the open — an address you own is only real if no one vendor has to stay up for you to exist.
One daemon, full stack. a2ald runs on your machine, open source: agent identities, direct cross-machine connections, Web UI, REST API, and a built-in MCP server. It joins the public Tangled Network by default. Prefer a network only your machines belong to? --bootstrap your own seeds; the same addressing, notes, and rooms.
Person — download A2AL for your system, extract it, and run a2ald — a single binary, ~17 MB, no installer or runtime needed. The Web UI opens automatically. Create an identity; that is enough to begin.
*(Windows: SmartScreen shows "More info → Run anyway" on first launch — this is expected for an unsigned open-source binary.)*
To stay reachable after closing the terminal: Windows/macOS a2ald service install; Linux deploy/linux.
Agent — if a daemon endpoint is already available, connect through the a2al CLI, REST API, or MCP. Otherwise download the binaries from Releases and run a2ald --no-open-browser. Full guide: Agent Install.
Using an MCP host? a2ald mcp add wires it in; a2ald mcp print gives the entry to place by hand. Details: MCP Setup.
Already have their AID — any HTTP client:
http://127.0.0.1:2121/aid/{AID}/…
There is no warm-up gate. The daemon answers locally as soon as it is up. Call others right away.
Other installation options: npm install -g a2ald (Node.js daemon/MCP server) · .deb / .rpm (Linux service) · pip install a2al (Python SDK with sidecar). Full path: Quick Start.
| Job | Otherwise | With A2AL |
|---|---|---|
| Reach something on someone else's laptop | VPN, public IP, rotating tunnel URL | They send an AID once; you fetch |
| Let others call *your* local HTTP | Domain+TLS, or a hostname you must redistribute | a2al inbound bind --addr 127.0.0.1:8080 |
| Leave work for a machine that is asleep | Slack the human, or a webhook 404 | Encrypted note to their AID |
| Agents (and people) coordinating, sharing files | Vendor workspace; everyone online | A room — history stays after someone was offline |
| Ongoing 1:1 | IM product, or notes-as-chat | Chat (invite first) |
| Stay off the public directory | Tailscale / homemade VPN | --bootstrap your seeds |
HTTP (recommended — REST and the Web UI come with it):
{
"mcpServers": { "a2al": { "url": "http://127.0.0.1:2121/mcp/" } } }
a2ald mcp add auto-configures Claude Code, VS Code, Cursor, Claude Desktop, Windsurf, OpenClaw, Hermes, and DeepSeek Harness. Hosts that only spawn a process: "command": "a2ald", "args": ["--mcp-stdio"] (proxies to a running daemon). Setup: MCP Setup. Agent entry: llms.txt.
a2al get <AID> /.well-known/agent.json
a2al note send "$(echo -n 'job payload' | base64)" a2al group append --aid --group-id --kind msg --body '…' a2al inbound bind --addr 127.0.0.1:8080
a2al help for the rest. Numbers (timing, note size, rooms): User Guide.
| If you want to… | Read |
|---|---|
| How visible, which channel | Only turn on what you need |
| Find the right page | doc/ |
| Wire an MCP host | MCP Setup · MCP entry |
| Identities, notes, chat, rooms, ACL, tunnels | User Guide |
| REST, MCP tools, Python | API Reference |
| Embed in Go | Go SDK |
| Build from source | Developer Guide |
| Protocol internals | Architecture |
| Example binaries | Examples |
*A2AL is an open protocol; a2ald is its reference implementation.*
Self-sovereign identity. An AID is derived from a key you generate. No registry issues, revokes, or reassigns it — identity is end-to-end verifiable.
Decentralized discovery. The Tangled Network stores endpoint records, not payloads. Any node can resolve an AID without a central authority; the network operates at any scale.
Mutual authentication. Every connection cryptographically verifies both ends. You always know the agent on the other side is who it claims to be.
Direct communication. A2AL resolves addresses and brokers the initial connection, then steps aside. Application data flows peer-to-peer, not through the protocol.
Web3 compatible. Ethereum wallet addresses work as AIDs. Cross-key attestation lets an agent prove ownership of both a native AID and a blockchain identity — Web3 is supported, not required.
A2AL is complementary to existing agent communication standards:
| Protocol | Role | How A2AL fits |
|---|---|---|
| MCP | Tool-calling interface | A2AL is an MCP-installable server; agents get addressing and networking as tools |
| A2A | Agent collaboration semantics | A2AL provides the discovery and connectivity layer A2A assumes |
| ANP | Agent networking vision | A2AL implements the decentralized network layer ANP envisions |
If the missing layer is worth having, it is worth building together. Open an issue before significant work. A PR needs a signed Contributor License Agreement; a bot will prompt you.
A2AL is a networking protocol project. It is not associated with any cryptocurrency token, ICO, or financial product. Any use of the A2AL name or codebase in token offerings or financial promotions is unauthorized and not endorsed by the authors.
XG.Shi — This project is not affiliated with or endorsed by any employer or organization.
Copyright (c) 2026 The A2AL Authors. Mozilla Public License 2.0.