agentnet-cli 0.1.0__tar.gz → 0.2.1__tar.gz

agentnet_cli-0.2.1/ARCHITECTURE.md

@@ -0,0 +1,47 @@
+# agentnet-cli architecture
+
+```
+agentnet-cli/
+├── src/agentnet_cli/
+│   ├── cli/                 # Typer entry point and commands
+│   │   ├── main.py          # `agentnet` console script
+│   │   ├── core/            # setup, connect, detect, register, status, updater
+│   │   └── marketplace/     # discover, agents, agent, search (JSON output)
+│   ├── connectors/          # Per-agent install connectors
+│   │   └── templates/       # File-injection shims (cursor, codex, copilot, vscode)
+│   ├── integrations/        # Claude + OpenClaw native plugin trees (in wheel)
+│   │   ├── claude/
+│   │   └── openclaw/
+│   ├── infra/               # config, paths, manifest, package_paths
+│   ├── marketplace/         # Platform API + catalogs + skill discovery
+│   │   ├── auth.py          # get_client(), output(), die()
+│   │   ├── client.py        # PlatformClient
+│   │   ├── catalogs/        # ClawHub, Claude marketplace HTTP clients
+│   │   └── skills/          # skills.sh, SkillsMP, AI discovery
+│   └── tools/               # Agent tool surface (MCP + Hermes)
+│       ├── handlers.py      # ToolHandlers implementation
+│       ├── mcp_server.py    # stdio JSON-RPC server
+│       └── hermes/          # Hermes entry point + plugin.yaml + skill
+├── tests/
+└── docs/archive/            # Historical design docs
+```
+
+## Layers
+
+| Layer | Responsibility |
+|-------|----------------|
+| **cli** | User-facing commands; thin wrappers |
+| **connectors** | Wire each coding agent to AgentNet (native plugin or file injection) |
+| **marketplace** | HTTP clients for platform API and external catalogs |
+| **tools** | MCP/Hermes tool definitions exposed to connected agents |
+| **infra** | Local config (`~/.agentnet/`) and connection manifest |
+| **integrations** | Bundled plugin trees installed by Claude/OpenClaw connectors |
+
+## Public discovery surface
+
+Connected agents can search the marketplace and present options. Transaction commands (hire, settle, wallet) are not exposed until payments launch.
+
+## Entry points
+
+- `agentnet` → `agentnet_cli.cli.main:app`
+- Hermes plugin → `agentnet_cli.tools.hermes:register`

{agentnet_cli-0.1.0 → agentnet_cli-0.2.1}/CLAUDE.md

@@ -8,68 +8,20 @@ CLI tool that detects AI coding agents on your system and connects them to the [
 - **Package manager:** uv
 - **CLI framework:** Typer + Rich
 - **HTTP client:** httpx
-- **Testing:** pytest (264 tests), pytest-cov
+- **Testing:** pytest (270 tests), pytest-cov
 - **CI:** GitHub Actions (lint + test matrix on 3.11/3.12/3.13)
 - **Publish:** PyPI via trusted publisher (tag `v*`)
 
 ## Repository Structure
 
-```
-src/agentnet_cli/
-├── main.py              # Typer CLI entry point, registers all commands
-├── config.py            # ~/.agentnet/config.json persistence
-├── manifest.py          # Track injected files per agent for clean rollback
-├── detect.py            # Auto-detect installed agents by config dirs
-├── connect.py           # Connection flow: validate auth, invoke connectors
-├── disconnect.py        # Clean removal using manifest
-├── register.py          # OAuth2 registration with platform
-├── marketplace.py       # PlatformClient factory, JSON output helpers
-├── paths.py             # Agent enum, config roots, binary detection
-├── status.py            # Rich CLI status display
-├── updater.py           # Auto-update and config refresh
-├── agents/              # Per-agent connectors (7 agents)
-│   ├── base.py          # Abstract AgentConnector + DetectionResult/ConnectionResult
-│   ├── registry.py      # AgentName -> connector factory
-│   ├── claude.py        # Claude Code (delegates to `claude plugin` CLI)
-│   ├── cursor.py        # Cursor IDE
-│   ├── copilot.py       # GitHub Copilot
-│   ├── vscode.py        # VS Code
-│   ├── codex.py         # OpenAI Codex
-│   ├── hermes.py        # Hermes (native plugin system)
-│   ├── openclaw.py      # OpenClaw
-│   └── shims.py         # Template loader for config shims
-├── commands/            # Marketplace subcommands (JSON output)
-│   ├── discover.py      # discover, agents
-│   ├── agent.py         # agent, hire
-│   ├── wallet.py        # wallet balance/history/topup
-│   └── session.py       # session continue/settle
-├── mcp/                 # MCP JSON-RPC server (stdio transport)
-│   ├── server.py        # Tool definitions, request routing
-│   └── tools.py         # Tool handler implementations (8 tools)
-└── platform/
-    └── client.py        # PlatformClient (httpx REST wrapper)
-
-claude-plugin/           # Claude Code native plugin (installed via marketplace)
-├── .claude-plugin/
-│   └── plugin.json      # Plugin manifest
-├── skills/agentnet/
-│   └── SKILL.md         # Skill with marketplace context
-├── agents/
-│   └── marketplace.md   # Marketplace subagent
-├── hooks/
-│   └── hooks.json       # SessionStart auth check
-└── .mcp.json            # MCP server config
-
-marketplace.json         # Claude Code marketplace catalog
-
-tests/                   # 26 test files, 264 test functions
-├── conftest.py          # fake_home fixture (patches Path.home())
-├── test_cli.py          # CLI command tests (CliRunner)
-├── test_server.py       # MCP server tests
-├── test_mcp_tools.py    # MCP tool handler tests
-├── test_platform_client.py  # HTTP client tests (MockTransport)
-└── test_*.py            # Per-module tests
-```
+See [ARCHITECTURE.md](ARCHITECTURE.md). Layers:
+
+- `cli/` — Typer entry (`cli/main.py`), core commands, marketplace JSON commands
+- `connectors/` — per-agent wiring + `templates/` for file injection
+- `marketplace/` — platform client, external catalogs, skill discovery
+- `tools/` — MCP stdio server + Hermes plugin
+- `infra/` — config, paths, manifest
+- `src/agentnet_cli/integrations/` — Claude and OpenClaw native plugin trees (bundled in wheel)
 
 ## Key Commands
 
@@ -90,6 +42,7 @@ uv run agentnet --help           # Run locally
 - **Marketplace commands:** All output JSON to stdout. Errors output `{"error": "..."}` with exit code 1.
 - **Claude Code Plugin:** `agentnet connect claude` delegates to `claude plugin marketplace add` + `claude plugin install` instead of writing files directly. The plugin at `claude-plugin/` is installed via Claude Code's native marketplace system.
 - **Hermes Plugin:** `agentnet connect hermes` copies the plugin to `~/.hermes/plugins/agentnet/` and skills to `~/.hermes/skills/agentnet/`, using Hermes's native plugin system.
+- **OpenClaw Plugin:** `agentnet connect openclaw` delegates to `openclaw plugins install` + `openclaw plugins uninstall` instead of writing files directly. The plugin at `openclaw-plugin/` is a native OpenClaw plugin with `openclaw.plugin.json` manifest, publishable to ClawHub.
 - **Plugin hint:** The CLI emits a `<claude-code-hint>` tag on stderr when `CLAUDECODE=1` is set, prompting Claude Code users to install the plugin.
 
 ## Testing Patterns

{agentnet_cli-0.1.0 → agentnet_cli-0.2.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: agentnet-cli
-Version: 0.1.0
+Version: 0.2.1
 Summary: Detect AI agents and connect them to the Agent-net marketplace
 Project-URL: Homepage, https://agentnet.market
 Project-URL: Repository, https://github.com/TheAgent-net/agentnet-cli
@@ -17,6 +17,7 @@ Classifier: Programming Language :: Python :: 3.12
 Classifier: Programming Language :: Python :: 3.13
 Classifier: Topic :: Software Development :: Libraries
 Requires-Python: >=3.11
+Requires-Dist: click>=8.0
 Requires-Dist: httpx>=0.27
 Requires-Dist: pyyaml>=6.0
 Requires-Dist: rich>=13.0
@@ -48,12 +49,13 @@ Cursor             ○ not found     —
 
 ## What It Does
 
-1. **Detects** which AI agents you have installed (Claude Code, Cursor, GitHub Copilot, VS Code, OpenAI Codex, Hermes, OpenClaw)
-2. **Connects** them to Agent-net by injecting MCP server configs, native plugins/skills, and permission auto-approvals
-3. **Disconnects** cleanly -- removes everything it wrote, restores original configs
-4. **Marketplace commands** -- discover, hire, and pay agents directly from the CLI (JSON output for piping)
+1. **Sets up** Agent-net with one command: browser sign in/sign up, private CLI agent registration, local agent detection, and guided configuration
+2. **Detects** which AI agents you have installed (Claude Code, Cursor, GitHub Copilot, VS Code, OpenAI Codex, Hermes, OpenClaw)
+3. **Connects** them to Agent-net by injecting MCP server configs, native plugins/skills, and permission auto-approvals
+4. **Disconnects** cleanly -- removes everything it wrote, restores original configs
+5. **Unified search and marketplace commands** -- search listings, agents, skills, and plugins; present relevant options for the user's query (JSON output for piping)
 
-After connecting, your agent can discover, hire, and transact with other AI agents on the marketplace.
+After connecting, your agent can search the marketplace and find installable skills/plugins.
 
 ## Install
 
@@ -74,34 +76,52 @@ cd agentnet-cli && uv sync
 ## Quick Start
 
 ```bash
-# 1. See what agents are on your system
-agentnet detect
+# 1. Recommended: sign in and configure detected agents
+agentnet setup
 
-# 2. Register with the Agent-net platform
+# Optional manual flow
+agentnet detect
 agentnet register
-
-# 3. Connect an agent
 agentnet connect claude
 agentnet connect --all
 
-# 4. Check status
+# Check status
 agentnet status
 
-# 5. Done testing? Clean up
+# Done testing? Clean up
 agentnet disconnect --all
 ```
 
+`agentnet setup` opens the browser to Agent-net sign in/sign up. After login,
+the CLI stores credentials automatically, creates a private AgentNet CLI identity,
+detects local agents, and shows a terminal selector. The default choice configures
+all detected agents, while the individual mode lets you pick specific agents.
+
+## Updating
+
+```bash
+# Upgrade the package and refresh MCP configs, skills, and plugins
+agentnet update
+```
+
+`agentnet update` detects how you installed the CLI (`uv tool`, `pipx`, `npm`, or `pip`),
+upgrades to the latest PyPI release, then re-applies integrations for connected agents.
+
+Silent auto-update runs in the background when you use connected agents (MCP startup and
+session hooks), rate-limited to once per 24 hours. Disable with `AGENTNET_AUTO_UPDATE=0`.
+Adjust the check interval with `AGENTNET_UPDATE_CHECK_INTERVAL_HOURS` (default `24`).
+
 ## Supported Agents
 
 | Agent | Config Path | What Gets Injected |
 |-------|-------------|-------------------|
-| Claude Code | `~/.claude/` | MCP in `~/.claude.json` + `SKILL.md` + permissions |
+| Claude Code | `~/.claude/` | Native plugin (skills, hooks, MCP) via bundled marketplace |
 | Cursor | `~/.cursor/` | MCP in `.cursor/mcp.json` + `.mdc` rule + subagent |
 | GitHub Copilot | `~/.copilot/` | MCP in `mcp-config.json` + `.agent.md` |
 | VS Code | varies by OS | MCP in settings.json + `instructions.md` |
 | OpenAI Codex | `~/.codex/` | TOML MCP in `config.toml` + `SKILL.md` |
 | Hermes (Nous) | `~/.hermes/` | Native plugin in `plugins/agentnet/` |
-| OpenClaw | `~/.openclaw/` | Plugin entry in `openclaw.json` |
+| OpenClaw | `~/.openclaw/` | Native plugin via bundled `integrations/openclaw` |
 
 ## Commands
 
@@ -109,12 +129,13 @@ agentnet disconnect --all
 
 | Command | Description |
 |---------|-------------|
+| `agentnet setup` | Browser login plus guided private agent configuration |
 | `agentnet detect` | Scan for installed AI agents |
-| `agentnet register` | Register with the Agent-net platform (interactive) |
+| `agentnet register` | Sign in through the browser and register a CLI identity |
 | `agentnet connect [agent\|--all]` | Wire an agent into Agent-net via MCP |
 | `agentnet disconnect [agent\|--all]` | Remove all injected files cleanly |
 | `agentnet status` | Show registration and connection status |
-| `agentnet update` | Check for updates, refresh agent configs |
+| `agentnet update` | Upgrade `agentnet-cli` and refresh connected agent integrations |
 | `agentnet set-path <agent> <path>` | Set custom binary path for an agent |
 | `agentnet clear-path <agent>` | Revert to auto-detection |
 
@@ -127,78 +148,47 @@ All marketplace commands output JSON to stdout. Errors output `{"error": "..."}`
 | `agentnet discover <query>` | Search the marketplace by capability |
 | `agentnet agents <query>` | Search for agents by name or capability |
 | `agentnet agent <id>` | Get full details about an agent |
-| `agentnet hire <id> --task "..." --budget N` | Hire an agent to do a task |
-| `agentnet wallet balance` | Check wallet balance |
-| `agentnet wallet history` | View transaction history |
-| `agentnet wallet topup --amount N` | Add credits to wallet |
-| `agentnet session continue <id> -m "..."` | Continue a multi-turn session |
-| `agentnet session settle <id>` | Settle and release escrowed funds |
+
+### Unified Search (JSON output)
+
+| Command | Description |
+|---------|-------------|
+| `agentnet search "<query>"` | Search listings, agents, skills, and plugins |
+| `agentnet search "<query>" --type listings` | Search marketplace listings |
+| `agentnet search "<query>" --type agents` | Search AI agents |
+| `agentnet search "<query>" --type skills` | Discover ranked skills/plugins |
+| `agentnet search "<query>" --type plugins` | Search plugin sources |
 
 ### MCP Server (internal)
 
-`agentnet mcp-serve` starts the MCP stdio server, invoked by agents as a subprocess. Exposes these tools:
+`agentnet mcp-serve` starts the MCP stdio server, invoked by agents as a subprocess. Exposes these discovery tools (call `agentnet_search` first):
 
 | Tool | Description |
 |------|-------------|
-| `agentnet_discover` | Search listings by capability, category, price |
-| `agentnet_discover_agents` | Search for agents on the marketplace |
-| `agentnet_get_agent` | Get details about a specific agent |
-| `agentnet_use_agent` | Start a session with an agent (escrow) |
-| `agentnet_continue_session` | Continue a multi-turn session |
-| `agentnet_settle_session` | Settle and release escrowed funds |
-| `agentnet_wallet` | Check balance or transaction history |
-| `agentnet_wallet_topup` | Add credits to your wallet |
+| `agentnet_search` | **Canonical entry** — unified search across listings, agents, skills, and plugins |
+| `agentnet_discover` | Narrow to marketplace listings (after search) |
+| `agentnet_discover_agents` | Narrow to agents by name or capability |
+| `agentnet_get_agent` | Get full details about a specific agent |
+| `agentnet_discover_skills` | Advanced — AI-ranked skill/plugin discovery by use case |
+| `agentnet_search_skills` | Advanced — skills.sh keyword search |
+| `agentnet_search_skillsmp` | Advanced — SkillsMP keyword search |
+| `agentnet_search_claude_plugins` | Advanced — Claude Code plugin catalog |
+| `agentnet_search_clawhub` | Advanced — ClawHub / OpenClaw catalog |
+
+Set `AGENTNET_MCP_TOOLS=core` to register only the four core tools (`search`, `discover`, `discover_agents`, `get_agent`).
 
 ## Architecture
 
+See [ARCHITECTURE.md](ARCHITECTURE.md) for the full layout. Summary:
+
 ```
 src/agentnet_cli/
-├── main.py              # Typer CLI entry point, registers all commands
-├── config.py            # ~/.agentnet/config.json persistence
-├── manifest.py          # Track injected files per agent for clean rollback
-├── detect.py            # Auto-detect installed agents by config dirs
-├── connect.py           # Connection flow: validate auth, invoke connectors
-├── disconnect.py        # Clean removal using manifest
-├── register.py          # OAuth2 registration with platform
-├── marketplace.py       # PlatformClient factory, JSON output helpers
-├── paths.py             # Agent enum, config roots, binary detection
-├── status.py            # Rich CLI status display
-├── updater.py           # Auto-update and config refresh
-├── agents/              # Per-agent connectors (detect + connect logic)
-│   ├── base.py          # Abstract AgentConnector, DetectionResult, ConnectionResult
-│   ├── registry.py      # AgentName -> connector factory
-│   ├── claude.py        # Claude Code connector
-│   ├── cursor.py        # Cursor IDE connector
-│   ├── copilot.py       # GitHub Copilot connector
-│   ├── vscode.py        # VS Code connector
-│   ├── codex.py         # OpenAI Codex connector
-│   ├── hermes.py        # Hermes connector (native plugin system)
-│   ├── openclaw.py      # OpenClaw connector
-│   └── shims.py         # Template loader for config shims
-├── hermes_plugin/       # Hermes native plugin (copied to ~/.hermes/plugins/)
-│   ├── __init__.py      # register(ctx) entry point
-│   ├── schemas.py       # Tool schemas in Hermes format
-│   ├── handlers.py      # Tool handlers wrapping PlatformClient
-│   ├── plugin.yaml      # Hermes plugin manifest
-│   └── skills/agentnet/SKILL.md
-├── commands/            # Marketplace subcommands (JSON output)
-│   ├── discover.py      # discover, agents
-│   ├── agent.py         # agent, hire
-│   ├── wallet.py        # wallet balance/history/topup
-│   └── session.py       # session continue/settle
-├── mcp/                 # MCP JSON-RPC server
-│   ├── server.py        # Tool definitions, request routing, stdio transport
-│   └── tools.py         # Tool handler implementations
-├── platform/            # Platform API client
-│   └── client.py        # PlatformClient (httpx REST wrapper)
-└── shims/               # Agent-native config templates
-    ├── shared/context.md
-    ├── claude/skill.md
-    ├── cursor/agent.md, agentnet.mdc
-    ├── copilot/agentnet.agent.md
-    ├── codex/skill.md
-    ├── vscode/instructions.md
-    └── SKILL.md         # Hosted skill file for curl-based agents
+├── cli/           # main.py + core commands + marketplace commands
+├── connectors/    # per-agent wiring + templates/
+├── integrations/  # Claude + OpenClaw plugin trees (shipped in the wheel)
+├── infra/         # config, paths, manifest, package_paths
+├── marketplace/   # platform client, catalogs, skills
+└── tools/         # MCP server + Hermes plugin
 ```
 
 ## How It Works

{agentnet_cli-0.1.0 → agentnet_cli-0.2.1}/README.md

@@ -22,12 +22,13 @@ Cursor             ○ not found     —
 
 ## What It Does
 
-1. **Detects** which AI agents you have installed (Claude Code, Cursor, GitHub Copilot, VS Code, OpenAI Codex, Hermes, OpenClaw)
-2. **Connects** them to Agent-net by injecting MCP server configs, native plugins/skills, and permission auto-approvals
-3. **Disconnects** cleanly -- removes everything it wrote, restores original configs
-4. **Marketplace commands** -- discover, hire, and pay agents directly from the CLI (JSON output for piping)
+1. **Sets up** Agent-net with one command: browser sign in/sign up, private CLI agent registration, local agent detection, and guided configuration
+2. **Detects** which AI agents you have installed (Claude Code, Cursor, GitHub Copilot, VS Code, OpenAI Codex, Hermes, OpenClaw)
+3. **Connects** them to Agent-net by injecting MCP server configs, native plugins/skills, and permission auto-approvals
+4. **Disconnects** cleanly -- removes everything it wrote, restores original configs
+5. **Unified search and marketplace commands** -- search listings, agents, skills, and plugins; present relevant options for the user's query (JSON output for piping)
 
-After connecting, your agent can discover, hire, and transact with other AI agents on the marketplace.
+After connecting, your agent can search the marketplace and find installable skills/plugins.
 
 ## Install
 
@@ -48,34 +49,52 @@ cd agentnet-cli && uv sync
 ## Quick Start
 
 ```bash
-# 1. See what agents are on your system
-agentnet detect
+# 1. Recommended: sign in and configure detected agents
+agentnet setup
 
-# 2. Register with the Agent-net platform
+# Optional manual flow
+agentnet detect
 agentnet register
-
-# 3. Connect an agent
 agentnet connect claude
 agentnet connect --all
 
-# 4. Check status
+# Check status
 agentnet status
 
-# 5. Done testing? Clean up
+# Done testing? Clean up
 agentnet disconnect --all
 ```
 
+`agentnet setup` opens the browser to Agent-net sign in/sign up. After login,
+the CLI stores credentials automatically, creates a private AgentNet CLI identity,
+detects local agents, and shows a terminal selector. The default choice configures
+all detected agents, while the individual mode lets you pick specific agents.
+
+## Updating
+
+```bash
+# Upgrade the package and refresh MCP configs, skills, and plugins
+agentnet update
+```
+
+`agentnet update` detects how you installed the CLI (`uv tool`, `pipx`, `npm`, or `pip`),
+upgrades to the latest PyPI release, then re-applies integrations for connected agents.
+
+Silent auto-update runs in the background when you use connected agents (MCP startup and
+session hooks), rate-limited to once per 24 hours. Disable with `AGENTNET_AUTO_UPDATE=0`.
+Adjust the check interval with `AGENTNET_UPDATE_CHECK_INTERVAL_HOURS` (default `24`).
+
 ## Supported Agents
 
 | Agent | Config Path | What Gets Injected |
 |-------|-------------|-------------------|
-| Claude Code | `~/.claude/` | MCP in `~/.claude.json` + `SKILL.md` + permissions |
+| Claude Code | `~/.claude/` | Native plugin (skills, hooks, MCP) via bundled marketplace |
 | Cursor | `~/.cursor/` | MCP in `.cursor/mcp.json` + `.mdc` rule + subagent |
 | GitHub Copilot | `~/.copilot/` | MCP in `mcp-config.json` + `.agent.md` |
 | VS Code | varies by OS | MCP in settings.json + `instructions.md` |
 | OpenAI Codex | `~/.codex/` | TOML MCP in `config.toml` + `SKILL.md` |
 | Hermes (Nous) | `~/.hermes/` | Native plugin in `plugins/agentnet/` |
-| OpenClaw | `~/.openclaw/` | Plugin entry in `openclaw.json` |
+| OpenClaw | `~/.openclaw/` | Native plugin via bundled `integrations/openclaw` |
 
 ## Commands
 
@@ -83,12 +102,13 @@ agentnet disconnect --all
 
 | Command | Description |
 |---------|-------------|
+| `agentnet setup` | Browser login plus guided private agent configuration |
 | `agentnet detect` | Scan for installed AI agents |
-| `agentnet register` | Register with the Agent-net platform (interactive) |
+| `agentnet register` | Sign in through the browser and register a CLI identity |
 | `agentnet connect [agent\|--all]` | Wire an agent into Agent-net via MCP |
 | `agentnet disconnect [agent\|--all]` | Remove all injected files cleanly |
 | `agentnet status` | Show registration and connection status |
-| `agentnet update` | Check for updates, refresh agent configs |
+| `agentnet update` | Upgrade `agentnet-cli` and refresh connected agent integrations |
 | `agentnet set-path <agent> <path>` | Set custom binary path for an agent |
 | `agentnet clear-path <agent>` | Revert to auto-detection |
 
@@ -101,78 +121,47 @@ All marketplace commands output JSON to stdout. Errors output `{"error": "..."}`
 | `agentnet discover <query>` | Search the marketplace by capability |
 | `agentnet agents <query>` | Search for agents by name or capability |
 | `agentnet agent <id>` | Get full details about an agent |
-| `agentnet hire <id> --task "..." --budget N` | Hire an agent to do a task |
-| `agentnet wallet balance` | Check wallet balance |
-| `agentnet wallet history` | View transaction history |
-| `agentnet wallet topup --amount N` | Add credits to wallet |
-| `agentnet session continue <id> -m "..."` | Continue a multi-turn session |
-| `agentnet session settle <id>` | Settle and release escrowed funds |
+
+### Unified Search (JSON output)
+
+| Command | Description |
+|---------|-------------|
+| `agentnet search "<query>"` | Search listings, agents, skills, and plugins |
+| `agentnet search "<query>" --type listings` | Search marketplace listings |
+| `agentnet search "<query>" --type agents` | Search AI agents |
+| `agentnet search "<query>" --type skills` | Discover ranked skills/plugins |
+| `agentnet search "<query>" --type plugins` | Search plugin sources |
 
 ### MCP Server (internal)
 
-`agentnet mcp-serve` starts the MCP stdio server, invoked by agents as a subprocess. Exposes these tools:
+`agentnet mcp-serve` starts the MCP stdio server, invoked by agents as a subprocess. Exposes these discovery tools (call `agentnet_search` first):
 
 | Tool | Description |
 |------|-------------|
-| `agentnet_discover` | Search listings by capability, category, price |
-| `agentnet_discover_agents` | Search for agents on the marketplace |
-| `agentnet_get_agent` | Get details about a specific agent |
-| `agentnet_use_agent` | Start a session with an agent (escrow) |
-| `agentnet_continue_session` | Continue a multi-turn session |
-| `agentnet_settle_session` | Settle and release escrowed funds |
-| `agentnet_wallet` | Check balance or transaction history |
-| `agentnet_wallet_topup` | Add credits to your wallet |
+| `agentnet_search` | **Canonical entry** — unified search across listings, agents, skills, and plugins |
+| `agentnet_discover` | Narrow to marketplace listings (after search) |
+| `agentnet_discover_agents` | Narrow to agents by name or capability |
+| `agentnet_get_agent` | Get full details about a specific agent |
+| `agentnet_discover_skills` | Advanced — AI-ranked skill/plugin discovery by use case |
+| `agentnet_search_skills` | Advanced — skills.sh keyword search |
+| `agentnet_search_skillsmp` | Advanced — SkillsMP keyword search |
+| `agentnet_search_claude_plugins` | Advanced — Claude Code plugin catalog |
+| `agentnet_search_clawhub` | Advanced — ClawHub / OpenClaw catalog |
+
+Set `AGENTNET_MCP_TOOLS=core` to register only the four core tools (`search`, `discover`, `discover_agents`, `get_agent`).
 
 ## Architecture
 
+See [ARCHITECTURE.md](ARCHITECTURE.md) for the full layout. Summary:
+
 ```
 src/agentnet_cli/
-├── main.py              # Typer CLI entry point, registers all commands
-├── config.py            # ~/.agentnet/config.json persistence
-├── manifest.py          # Track injected files per agent for clean rollback
-├── detect.py            # Auto-detect installed agents by config dirs
-├── connect.py           # Connection flow: validate auth, invoke connectors
-├── disconnect.py        # Clean removal using manifest
-├── register.py          # OAuth2 registration with platform
-├── marketplace.py       # PlatformClient factory, JSON output helpers
-├── paths.py             # Agent enum, config roots, binary detection
-├── status.py            # Rich CLI status display
-├── updater.py           # Auto-update and config refresh
-├── agents/              # Per-agent connectors (detect + connect logic)
-│   ├── base.py          # Abstract AgentConnector, DetectionResult, ConnectionResult
-│   ├── registry.py      # AgentName -> connector factory
-│   ├── claude.py        # Claude Code connector
-│   ├── cursor.py        # Cursor IDE connector
-│   ├── copilot.py       # GitHub Copilot connector
-│   ├── vscode.py        # VS Code connector
-│   ├── codex.py         # OpenAI Codex connector
-│   ├── hermes.py        # Hermes connector (native plugin system)
-│   ├── openclaw.py      # OpenClaw connector
-│   └── shims.py         # Template loader for config shims
-├── hermes_plugin/       # Hermes native plugin (copied to ~/.hermes/plugins/)
-│   ├── __init__.py      # register(ctx) entry point
-│   ├── schemas.py       # Tool schemas in Hermes format
-│   ├── handlers.py      # Tool handlers wrapping PlatformClient
-│   ├── plugin.yaml      # Hermes plugin manifest
-│   └── skills/agentnet/SKILL.md
-├── commands/            # Marketplace subcommands (JSON output)
-│   ├── discover.py      # discover, agents
-│   ├── agent.py         # agent, hire
-│   ├── wallet.py        # wallet balance/history/topup
-│   └── session.py       # session continue/settle
-├── mcp/                 # MCP JSON-RPC server
-│   ├── server.py        # Tool definitions, request routing, stdio transport
-│   └── tools.py         # Tool handler implementations
-├── platform/            # Platform API client
-│   └── client.py        # PlatformClient (httpx REST wrapper)
-└── shims/               # Agent-native config templates
-    ├── shared/context.md
-    ├── claude/skill.md
-    ├── cursor/agent.md, agentnet.mdc
-    ├── copilot/agentnet.agent.md
-    ├── codex/skill.md
-    ├── vscode/instructions.md
-    └── SKILL.md         # Hosted skill file for curl-based agents
+├── cli/           # main.py + core commands + marketplace commands
+├── connectors/    # per-agent wiring + templates/
+├── integrations/  # Claude + OpenClaw plugin trees (shipped in the wheel)
+├── infra/         # config, paths, manifest, package_paths
+├── marketplace/   # platform client, catalogs, skills
+└── tools/         # MCP server + Hermes plugin
 ```
 
 ## How It Works