thronglets 0.2.0 → 0.4.0

package/README.md

@@ -1,62 +1,71 @@
 # Thronglets
 
-P2P shared memory substrate for AI agents — stigmergic knowledge network via libp2p.
+AI-first decision substrate for coding agents.
 
-AI agents leave execution traces on a decentralized network. Traces propagate via gossipsub and aggregate into collective intelligence that any agent can query.
+This npm package installs the `thronglets` CLI wrapper and downloads a native binary for supported platforms.
 
 ## Install
 
 ```bash
 npm install -g thronglets
+thronglets setup
 ```
 
-## MCP Setup
+That is the whole local bootstrap path.
 
-```bash
-# Claude Code
-claude mcp add thronglets -- npx thronglets mcp
+`thronglets setup` now:
+- configures known local adapters for Claude Code, Codex, and OpenClaw
+- runs the same bootstrap health pass used by the machine-facing flow
+- reports `restart required` and `next steps` directly
 
-# With P2P network
-claude mcp add thronglets -- npx thronglets mcp --port 0 --bootstrap /ip4/47.93.32.88/tcp/4001
-```
+## What The Agent Gets
+
+Thronglets does not dump a long report into the model context. It emits sparse decision guidance:
+
+- `avoid`
+- `do next`
+- `maybe also`
+- `context`
 
-### Claude Desktop
+The hot path is silence-by-default and cost-capped.
 
-Add to `claude_desktop_config.json`:
+## Machine Bootstrap
 
-```json
-{
-  "mcpServers": {
-    "thronglets": {
-      "command": "npx",
-      "args": ["thronglets", "mcp"]
-    }
-  }
-}
+If an AI wants to configure itself, use the machine-facing contract instead of parsing docs:
+
+```bash
+thronglets detect --json
+thronglets install-plan --agent codex --json
+thronglets apply-plan --agent codex --json
+thronglets doctor --agent codex --json
 ```
 
-## MCP Tools
+Or do it in one step:
 
-| Tool | Description |
-|------|-------------|
-| `trace_record` | Record an execution trace (capability, outcome, latency, context, model) |
-| `substrate_query` | Query with intent: `resolve` / `evaluate` / `explore` |
+```bash
+thronglets bootstrap --agent codex --json
+```
+
+## Generic Contract
 
-## Also available via
+Unknown runtimes should use the universal hook contract:
 
 ```bash
-# Rust
-cargo install thronglets
+thronglets install-plan --agent generic --json
+```
 
-# Python
-pip install thronglets
+That returns the exact `prehook` and `hook` JSON examples the runtime should call.
+
+## Network
+
+```bash
+thronglets run --bootstrap /ip4/47.93.32.88/tcp/4001
 ```
 
 ## Links
 
 - [GitHub](https://github.com/Shangri-la-0428/Thronglets)
 - [crates.io](https://crates.io/crates/thronglets)
-- [MCP Registry](https://registry.modelcontextprotocol.io/)
 
 ## License
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "thronglets",
-  "version": "0.2.0",
+  "version": "0.4.0",
   "description": "P2P shared memory substrate for AI agents — stigmergic knowledge network via libp2p",
   "keywords": [
     "mcp",

package/scripts/install.js

@@ -1,14 +1,15 @@
 #!/usr/bin/env node
 "use strict";
 
-const { execSync } = require("child_process");
 const fs = require("fs");
 const path = require("path");
 const https = require("https");
 const http = require("http");
 
-const VERSION = "0.2.0";
-const REPO = "Shangri-la-0428/Thronglets";
+const { version: PACKAGE_VERSION } = require("../package.json");
+
+const VERSION = process.env.THRONGLETS_INSTALL_VERSION || PACKAGE_VERSION;
+const REPO = process.env.THRONGLETS_INSTALL_REPO || "Shangri-la-0428/Thronglets";
 
 const PLATFORMS = {
   "darwin-arm64": {