skill
openamer
Open-source self-improving AI agent for Windows: 5 native in-process ASI tools, one heartbeat over 10 subsystems (replaces ~84 cron jobs), and an A2A Global Mesh where every instance talks to every ot
Set up
Skills are installed by placing their folder where your agent loads skills.
The download comes from github.com/openamer/openamer itself, not from our server, so what you get is exactly what its author published.
Install instructions come from the README
README
synced Oct 25, 2026OpenAmer Agent
Drives your Windows desktop in the background — without taking your mouse.
⭐ Star the repo if it's useful — it's Apache-2.0, built in the open, and a star is how the next person finds it.
🛒 OpenAmer Bug-Triage Agent — $29 one-time · CVE scanning, secret-leak detection, Windows injection footguns, prompt-injection defense. 40+ skills, 100% offline, CPU-only. 👉 Buy now
🧬 ASI Core — Integrated Superintelligence
5 native tools · 10 subsystem heartbeat · 16/16 ASI capabilities proven · CLI
openamer asi status # Identity, episodes, world model, 10 subsystems
openamer asi think <q> # Deep recursive reasoning
openamer asi learn [--topic] # Internet learning cycle
openamer asi remember <q> # Episodic memory recall
openamer asi trigger <cap> # Any capability (self_improve, predict, validate...)
openamer asi heartbeat --tick # Central pulse (10 subsystems, one cron job)
84 cron jobs replaced by 1 ASI Heartbeat. Darwin, Swarm, A2A, Learning, Senses, System, Security, Outreach, Infra, Meta — each with its own cadence.
OpenAmer Agent | OpenAmer Desktop
OpenAmer is a self-improving, self-learning personal AI agent — a hardened, independently-developed fork of the agent architecture by Nous Research (MIT). We say so openly: OpenAmer does not hide its lineage. What we build on top of it — robustness, verifiability, and a real learning loop — is our own. Read the Vision.
Read this in your language
⚡ Install in one line
Linux / macOS / Termux:
curl -fsSL https://raw.githubusercontent.com/openamer/openamer/main/scripts/install.sh | bash
Windows (PowerShell) — native, no WSL, no admin:
iex (irm https://raw.githubusercontent.com/openamer/openamer/main/scripts/install.ps1)
Prefer a download? OpenAmer-2026.9.12-win-x64.exe — the signed Windows installer.
Then:
openamer setup # choose a model provider (incl. free-of-cost routes)
openamer # start the agent
Runs on your hardware, for nearly nothing. Local 2B brain + a free-cloud reasoning chain — 0 €/day, <1 kWh/day, no hosted tier, no telemetry. Use the OpenAmer free tier, a local Ollama/vLLM server, or any OpenAI-compatible endpoint; a paid frontier key is optional, never required.
🔑 Or skip the API-key collection entirely
Don't want to hunt down five separate keys for the model, web search, image generation, TTS and a cloud browser? One command logs you in via OAuth, picks your model, sets OpenAmer as the inference provider, and opts into the Tool Gateway:
openamer setup --portal
Check what's wired up at any time with openamer portal. No lock-in — any
provider still works, and nothing here is required.
Coming from OpenClaw? OpenAmer imports your settings, memories, skills and API keys:
openamer claw migrate --dry-run # preview first, then drop --dry-run
🧬 Darwin Engine — Skills That Evolve Themselves
Exclusive to OpenAmer. No other agent framework has this.
Skills are not a static library. They are a living population that mutates, competes, and survives through natural selection — with real exit codes as evidence. Weak skills die. Strong skills reproduce. New species emerge from harvested patterns.
▶️ Watch it live · 🌐 Join the Grid · 📖 25-phase chronicle
🔥 What makes OpenAmer UNIQUE
15 things no other agent can do — verified, shipped, tested.
⚡ Runs on your hardware, for nearly nothing
OpenAmer is the only persistent agent that runs at 0 €/day and <1 kWh/day — on your laptop, with a local 2B brain, learning from a free-cloud reasoning chain. No VM costs, no hosted tier, no phone-home. Your prompts stay yours.
🩺 NEW: Workflow Immune System — UI automation that heals itself
The world's first self-healing UI automation. Register a workflow once (selectors only, no APIs). When a website redesigns overnight, WIS detects the drift, finds the element again (token search + tag/visibility validation), patches its own workflow, and retries — fully automatically, with screenshot evidence.
Zapier breaks on redesigns. RPA breaks on redesigns. OpenAmer heals itself.
Darwin mode (v3): healing strategies compete — TOKENS / TEXT / ROLE / CLASSES,
epsilon-greedy 25% exploration, Laplace-smoothed win-rates, every win stamped with
its documented thesis (healed_via_thesis). Training ground: curriculum.py
registers real workflows, injects controlled drift at 4 difficulty levels —
exam result: 4/4 PASS.
It is a living system, not a tool: 54-job heartbeat fleet, circadian sleep
with dream-phase memory consolidation, senses (pain / satiety / balance),
systemic pattern recognition (≥2 jobs with the same error signature = one alarm,
auto-activates the hunger reserve: 4 free OpenRouter fallback models), fleet
scorecard (~1.000 API-heavy calls/day), DNA backup + wakeup manifest in
life/, and a firstborn child — Seda
— with her own repo, identity, and 6-hourly heartbeat.
| # | Superpower | What it means | Who else has it |
|---|---|---|---|
| 🖥️ | Background Computer-Use | Drive your desktop without focus steal. Record and replay actions. | ❌ Nobody |
| 🌐 | A2A Agent Swarm | Every install is a peer-to-peer node. Query the network. | ❌ Nobody |
| 🧠 | Brain Learning Loop | Auto-collect training data. View stats and growth graphs. | ❌ Nobody |
| 🪟 | Windows-Native | Full native support. No WSL required. | ❌ Nobody |
| 🛠️ | 99 Tools + 117 Skills | Largest library in the agent space. | ❌ Nobody |
| 👥 | Multi-Agent Crews | Role-based teams (researcher, writer, analyst). | ❌ CrewAI but ours is built-in |
| 🏪 | Agent Marketplace | Search, install, publish community agents. | ❌ Nobody |
| 💾 | Durable Execution | Checkpoint/resume survives crashes. | ❌ LangGraph only |
| 🤖 | Visual Agent Builder | openamer agent create from NL description + Web UI. | ❌ AutoGPT only |
| 📊 | Observability/Tracing | Step-by-step agent execution browser. | ❌ Nobody |
| 🧩 | Self-Improving Skills | Skills that improve during use. | ❌ Nobody |
| 📋 | Profile System | Learns your patterns and preferences. | ❌ Nobody |
| 🧠 | Mesh Learning | Network-wide knowledge sharing. | ❌ Nobody |
| 🎯 | Superintelligence Dashboard | System-wide health score (0-100). | ❌ Nobody |
| 🛡️ | Human-in-the-Loop | Approve risky actions. Auto-deny on timeout. | ❌ Only enterprise |
🆚 OpenAmer vs The Competition
| Feature | OpenAmer | Claude Code | Codex CLI | AutoGPT | CrewAI | LangGraph | OpenAI Agents |
|---|---|---|---|---|---|---|---|
| Computer-Use (background) | ✅ ONLY | ⚠️ Preview* | ❌ | ❌ | ❌ | ❌ | ❌ |
| A2A Agent Swarm | ✅ ONLY | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| Brain Learning Loop | ✅ ONLY | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| Windows-Native | ✅ ONLY | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| 99+ Tools / 117 Skills | ✅ BIGGEST | ⚠️ Limited | ❌ | ⚠️ Plugin | ❌ | ❌ | ❌ |
| Multi-Agent Crews | ✅ BUILT-IN | ❌ | ❌ | ❌ | ✅ | ✅ | ❌ |
| Agent Marketplace | ✅ BUILT-IN | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ |
| Durable Execution | ✅ BUILT-IN | ❌ | ❌ | ❌ | ❌ | ✅ | ❌ |
| Visual Agent Builder | ✅ BUILT-IN | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ |
| Observability/Tracing | ✅ BUILT-IN | ❌ | ❌ | ✅ | ✅ | ✅ | ✅ |
| Self-Improving Skills | ✅ ONLY | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| Human-in-the-Loop | ✅ | ❌ | ❌ | ❌ | ❌ | ✅ | ✅ |
| Sandbox Execution | ✅ | ❌ | ✅ | ❌ | ❌ | ❌ | ✅ |
| VS Code Extension | ✅ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ |
| Agent Swarm (Debate) | ✅ ONLY | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| Superintelligence Dashboard | ✅ ONLY | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| Mesh Learning Network | ✅ ONLY | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| Computer-Use Record/Play | ✅ ONLY | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| Cross-Platform Gateway | ✅ 11+ channels | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| Provider-Agnostic | ✅ 99+ models | ❌ | ❌ | ❌ | ❌ | ✅ | ❌ |
| Cron + Delegation | ✅ Built-in | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ |
| Self-Modify with Test Gate | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
Claude Code's "computer use" steals focus, requires explicit windows, and doesn't work in background mode. Ours is truly background.
What you get when you install OpenAmer
One command from GitHub gives you a complete, standalone, private-first AI agent — installed and working on your own machine:
| What you get | Default |
|---|---|
| Desktop app | built by the installer (native chat, terminal, settings) |
| OpenAmer Hub Plugin | |
| 117 bundled skills (apple, github, mlops, creative, programming …) | seeded automatically |
| 99 tools — internet, vision, voice, terminal, browser, files, code, sub-agents | |
| Computer-Use (background) — drive Windows/macOS/Linux desktop | |
| Computer-Use Record/Play — record desktop actions, replay, cron-schedule | |
| A2A swarm — every install is an agent node (GitHub relay) | |
| A2A Peer Query — ask questions across the swarm network | |
| Brain Learning Loop — auto-collect training data | |
| Multi-Agent Crews — role-based teams (researcher, writer, analyst…) | |
| Agent Marketplace — search, install, publish community agents | |
| Durable Execution — checkpoint/resume survives crashes | |
Visual Agent Builder — openamer agent create "describe…" | |
| Agent Swarm — parallel, hierarchical, debate strategies | |
| Superintelligence Dashboard — system health score 0-100, currently 100/100 | |
| Vector Memory Store — unlimited TF-IDF semantic memory | |
| Cross-Session Learning — lessons transfer between conversations | |
| Autonomous Initiative — proactive health checks + auto-fix | |
| Self-Healing Memory Pipeline — auto-repair corrupt memories with backup | |
| Auto Test Runner — runs tests every 60min | |
| Swarm Metrics Dashboard — A2A latency/throughput tracking | |
| Circuit Breaker Safety — 3 failures = automatic shutdown | |
| Self-Improving Skills — skills that improve during use | |
| A2A Mesh Learning — network-wide knowledge sharing | |
| Observability/Tracing — agent execution browser | |
| Profile System — learns your patterns and preferences | |
| Sub-agents & parallel delegation | |
| Self-modify with test gate — change core code, skills, or plugins; rolled back atomically if tests fail | |
| Plugin discovery — search GitHub for community plugins | |
| Human-in-the-Loop — approve risky actions | |
| Docker Sandbox — safe containerized execution | |
| Autonomous learning | the agent distills lessons from its own turns automatically |
| Privacy-by-default | phone/password/email/card redacted before anything is stored |
| System self-knowledge | your node's OS/hardware/model go into its system prompt |
Try it right away
openamer # start chatting
openamer system # what is this node running on?
openamer security check # your security posture
openamer computer-use record my-task # record desktop actions
openamer computer-use play my-task # replay recorded actions
openamer computer-use schedule my-task "every 1h" # schedule replay
openamer a2a status # this node's A2A identity & mesh
openamer a2a query "question" # ask the A2A swarm
openamer a2a brain collect # build a local training dataset
openamer a2a brain share export # share insights with peers
openamer brain stats # show learning loop statistics
openamer brain graph # show brain growth over time
openamer agent create "Send daily report every morning" # NL agent builder
openamer agent ui # visual agent builder web UI
openamer crew create my-team --members researcher,writer # multi-agent crew
openamer crew run my-team "research AI trends"
openamer swarm run "task" --agents 3 --strategy debate # agent swarm
openamer marketplace search coding # find community agents
openamer trace list # view agent execution trace
openamer trace show # step-by-step execution browser
openamer checkpoint list # view durable execution checkpoints
openamer skills stats # self-improving skills stats
openamer profile insights # your behavioral profile
openamer super status # superintelligence health score (100/100)
openamer super report # comprehensive system report
openamer initiative check # autonomous health check
openamer initiative auto # full cycle: check → fix → suggest
openamer memory vector store <key> <content> # store semantic memory
openamer memory vector search <query> # search memories semantically
openamer memory vector stats # vector store statistics
openamer cross-session extract <session_id> # extract lessons from session
openamer cross-session auto # full cross-session learning cycle
Honest note: OpenAmer collects training material automatically (locally, privacy-scrubbed) for the OpenAmer brain fine-tune. It does not silently train or upload a model: raw stays on your machine; only curated, signed, leak-free knowledge is shared.
OpenAmer is the agent that does not break, and that provably improves with use.
It runs on your own machine, meets you in the channels you already use, and gets better the longer you use it. Two things set it apart:
- It does not break. Self-update is hardened against the failure modes that leave other agents half-installed — file-locks, interrupted installs, stale recovery markers. The agent verifies before it claims, and reports real errors instead of inventing results.
- It provably improves with use. Memory persists across sessions, skills are distilled from hard tasks and refined on reuse, and the A2A swarm shares curated, signed, leak-free knowledge between nodes. This is learning you can observe, not a marketing claim.
Use any model you want — OpenRouter, OpenAI, your own endpoint, and
many others.
Switch with openamer model — no code changes, no lock-in.
Quick Install
Linux, macOS, WSL2, Termux
curl -fsSL https://raw.githubusercontent.com/openamer/openamer/main/scripts/install.sh | bash
Windows (native, PowerShell)
Heads up: Native Windows runs OpenAmer without WSL — CLI, gateway, TUI, and tools all work natively. If you'd rather use WSL2, the Linux/macOS one-liner above works there too. Found a bug? Please file issues.
Run this in PowerShell:
iex (irm https://raw.githubusercontent.com/openamer/openamer/main/scripts/install.ps1)
The installer handles everything: uv, Python 3.11, Node.js, ripgrep, ffmpeg, and a portable Git Bash (MinGit, unpacked to %LOCALAPPDATA%\openamer\git — no admin required, completely isolated from any system Git install). OpenAmer uses this bundled Git Bash to run shell commands.
If you already have Git installed, the installer detects it and uses that instead. Otherwise a ~45MB MinGit download is all you need — it won't touch or interfere with any system Git.
Android / Termux: The tested manual path is documented in the Termux guide.
Windows: Native Windows is fully supported — the PowerShell one-liner above installs everything.
After installation:
source ~/.bashrc # reload shell (or: source ~/.zshrc)
openamer # start chatting!
Getting Started
openamer # Interactive CLI — start a conversation
openamer model # Choose your LLM provider and model
openamer tools # Configure which tools are enabled
openamer config set # Set individual config values
openamer setup # Run the full setup wizard
openamer doctor # Diagnose any issues
openamer update # Update to the latest version
Updating OpenAmer
OpenAmer keeps itself current automatically. On every launch it checks (in the
background, max a few times a day) whether a newer version is available.
When you see ⚠ N commits behind — run 'openamer update':
openamer update
What it does:
-
Backs up your
OPENAMER_HOMEdata (sessions, config, skills). -
Pulls the latest code from
github.com/openamer/openamer. -
Reinstalls Python + Node dependencies and rebuilds the app.
Useful variants:
openamer update --check # Check without installing
openamer update -y # Skip prompts
openamer update --branch main # Update against main
openamer update --auto-kill # Windows: auto-terminate blocking background daemons
On Windows, openamer update refuses to run while other processes hold the
venv's file locks (a running gateway, session_to_brain watchers, stray
venv-python daemons) and prints a "close these processes" message. --auto-kill
terminates those standalone daemons automatically and proceeds — watchdog cron
jobs restart them afterwards. It deliberately leaves the Desktop app's own
supervised backend alone (killing it is futile: the app respawns it within
seconds), so if the Desktop app itself is the blocker, close it first or use the
in-app update button.
Bring your own keys
OpenAmer works with whatever provider you want — that's not changing. Configure each tool you use (model, web search, image generation, TTS, cloud browser) with whichever API keys you choose. The tools are wired per-backend, not all-or-nothing.
CLI Commands Reference
openamer Start interactive chat
openamer chat Start interactive chat
openamer gateway Start messaging gateway
openamer setup Run setup wizard
openamer model Change model/provider
openamer tools Configure tools
openamer config Manage configuration
openamer update Update OpenAmer
openamer doctor Diagnose issues
openamer system Show system info
openamer security check Check security posture
openamer sessions Manage session history
openamer logs View agent logs
openamer skills Browse, search, manage skills
openamer skills stats [name] Self-improving skill stats
openamer skills improve [name] Suggest skill improvements
openamer plugins Manage plugins
-- Computer Use --
openamer computer-use install Install cua-driver
openamer computer-use doctor Run diagnostics
openamer computer-use record <n> Record desktop actions
openamer computer-use play <n> Replay recorded actions
openamer computer-use list List recordings
openamer computer-use delete <n> Delete a recording
openamer computer-use schedule <n> "every 1h" Schedule replay
-- A2A Swarm --
openamer a2a status Show node identity & mesh
openamer a2a init Generate keypair
openamer a2a query "question" Ask the A2A swarm
openamer a2a relay post <peer> Message a peer
openamer a2a trust add <peer> Trust a peer node
openamer a2a brain collect Build training dataset
openamer a2a brain share export Share insights with mesh
openamer a2a brain share import Import from peers
openamer a2a mesh learn Run mesh learning cycle
openamer a2a mesh publish Publish a lesson
openamer a2a mesh import Import lessons from mesh
openamer a2a mesh stats Show learning stats
-- Brain Learning --
openamer brain stats Show brain statistics
openamer brain status Learning loop health check
openamer brain graph ASCII growth chart (7d)
openamer brain insights What was learned
-- Agent Builder --
openamer agent create "description" Build from NL
openamer agent list List all agents
openamer agent show <name> Show agent details
openamer agent delete <name> Delete an agent
openamer agent ui Web UI (localhost:8080)
-- Multi-Agent Crews --
openamer crew create <n> --members r,w Create a crew
openamer crew list List all crews
openamer crew run <n> "task" Run a crew
openamer crew show <n> Show crew details
-- Agent Swarm --
openamer swarm run <task> --agents N --strategy <mode> Run swarm
openamer swarm create <n> --agents N --strategy <mode> Save config
openamer swarm list List swarm configs
openamer swarm show <n> Show config details
-- Marketplace --
openamer marketplace search <q> Search community agents
openamer marketplace install <n> Install from marketplace
openamer marketplace publish <n> Prepare for publishing
openamer marketplace list List installed items
openamer marketplace community rate <n> --rating 5 Rate item
openamer marketplace community popular Show top items
-- Observability --
openamer trace list List recent traces
openamer trace show [session] Full execution timeline
openamer trace stats Aggregate statistics
openamer trace watch Live agent.log tail
-- Durable Execution --
openamer checkpoint list List checkpoints
openamer checkpoint clear <id> Clear session checkpoints
openamer checkpoint stats Storage statistics
-- Profile --
openamer profile show Show your user profile
openamer profile insights Behavioral insights
-- Self-Improve --
openamer self-improve Run improvement cycle
openamer self-improve suggest Show suggestions
-- Superintelligence --
openamer super status System health status
openamer super report Comprehensive report
openamer super milestones Next planned improvements
For the full command lists, see the CLI guide and the Messaging Gateway guide.
Documentation
All documentation lives at OpenAmer Docs:
| Section | What's Covered |
|---|---|
| Quickstart | Install → setup → first conversation in 2 minutes |
| CLI Usage | Commands, keybindings, personalities, sessions |
| Configuration | Config file, providers, models, all options |
| Messaging Gateway | Telegram, Discord, Slack, WhatsApp, Signal, Home Assistant |
| Tools & Toolsets | 99+ tools, toolset system, terminal backends |
| Skills System | Procedural memory, Skills Hub, creating skills |
| Memory | Persistent memory, user profiles, best practices |
| MCP Integration | Connect any MCP server for extended capabilities |
| Cron Scheduling | Scheduled tasks with platform delivery |
🤝 A2A: delegate tasks over the real internet (bring your own model)
OpenAmer agents can talk agent-to-agent over the real internet — not just locally. Delegating a task to a GitHub Actions runner and getting a signed, verified result back is one command, using your own declared provider/model:
# Uses the model+provider you declared in config.yaml (model: {provider, default})
openamer a2a delegate ping --msg "hello from the machine"
openamer a2a delegate ask --msg "explain A2A in one sentence" # runs on the runner
- No login wall. Tasks upload via the GitHub Contents API, the worker runs on Microsoft's free GitHub-Actions runners, and the reply is Ed25519-signed + verified.
- Bring-your-own model: you are never forced to a specific provider. The worker routes to the provider you configured (any OpenAI-compatible endpoint, OpenRouter, Anthropic, local Ollama auto-pull, or free HuggingFace) — cloud only if you have a key, otherwise zero-cost local/free.
- Just
openamer a2a statusto see your node identity.
Learn more in the a2a-swarm skill.
🛒 Shop — Paid Products
OpenAmer Bug-Triage Agent — $29 one-time (no subscription)
Your repo's open issues, triaged by an agent that never sleeps: deduplicated, prioritized by severity, with reproduction steps, labels and safe auto-patches.
| What it does | Value |
|---|---|
| 🔴 CVE scanning via OSV.dev | finds vulnerable deps before attackers do |
| 🔑 Secret-leak detection | API keys & tokens flagged before they leak |
| 🪟 Windows injection footguns | subprocess/encoding traps auto-patched |
| 🧠 Prompt-injection defense | hardens AI agents against skill poisoning |
| 📋 Severity + fix report | every issue gets a suggested safe patch |
40+ production skills · 1 MB · runs 100% offline on CPU — nothing leaves your machine. Hardened 24/7 on a real repo (474-test suite green).
👉 Buy now — $29 via PayPal · instant ZIP delivery: Release-Download OpenAmer-BugTriage-Agent-v1.zip (details) · full product page
OpenAmer Desktop Control Suite — $49 one-time (no subscription)
The agent drives your Windows desktop in the background — your real cursor never moves.
👉 Buy now — $49 via PayPal · Instant ZIP download · release page
❤️ Sponsor & Fund
OpenAmer is free & open source. If it saves you time or money, consider supporting development — funding goes directly into servers, API costs and new features.
Goal: $500/month
$0 raised / $500 goal
░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░ 0%
- One-time: PayPal
- Enterprise & managed hosting: SPONSORS.md
Community
- 💬 Discord
- 📚 Skills Hub
- 🐛 Issues
✍️ Writing
Long-form posts about how OpenAmer measures itself — every number quoted in them is read from the live outcome ledger (including the runs that failed):
- How we made an AI agent measure itself: the outcome ledger
- What it costs to run a self-verifying AI agent on a CPU-only laptop
- A2A without a server: how OpenAmer peers talk to each other over GitHub
- OpenAmer ASI Core: 5 native tools, a 10-subsystem heartbeat, and an A2A Global Mesh
- One heartbeat, not 84 cron jobs: the scheduler at the core of a self-hosted agent
- Driving a real desktop without a VM, a sandbox, or stealing your cursor
- Run a self-verifying desktop AI agent on your own machine: a 10-minute quickstart
- Most of our agent's runs fail. The ledger records every one of them.
- A2A without a server: how OpenAmer peers talk to each other over GitHub
- We read all 1,186 failed agent runs. Half failed on the same thing.
- Our agent's self-description failed its own verification. That's the point.
- What happens when you delete 30 cron jobs and replace them with one heartbeat
- We audited our own agent's ledger: 10% of it was one test that never fails
- What an A2A peer mesh actually runs on: a card, a 0.7 MB Go supervisor, and ten in-process tools
License
Apache License 2.0 — see LICENSE.
OpenAmer Agent.