engrm 0.4.37 → 0.4.39

package/README.md

@@ -75,7 +75,7 @@ openclaw plugins install engrm-openclaw-plugin
 npx engrm init
 ```
 
-This auto-configures MCP servers and hooks in `~/.claude.json` and `~/.codex/config.toml`.
+This auto-configures MCP servers and hooks in Claude Code and Codex, and also sets up OpenCode with a local Engrm plugin plus MCP registration.
 
 **Alternative methods:**
 ```bash
@@ -91,6 +91,54 @@ npx engrm init --manual
 
 For npm users, Engrm runs on Node.js 18+ and does not require Bun to be installed.
 
+### For Hermes / Remote MCP
+
+Engrm can now run over Streamable HTTP for Hermes-style container deployments.
+
+Add this to `~/.engrm/settings.json`:
+
+```json
+{
+  "http": {
+    "enabled": true,
+    "port": 3767,
+    "bearer_tokens": ["replace-with-a-long-random-token"]
+  },
+  "fleet": {
+    "project_name": "shared-experience",
+    "namespace": "ns_fleet_shared",
+    "api_key": "cvk_fleet_shared"
+  }
+}
+```
+
+Then start the remote MCP endpoint:
+
+```bash
+ENGRM_HTTP_PORT=3767 engrm serve --http
+```
+
+Hermes should connect with a bearer token header:
+
+```json
+{
+  "mcpServers": {
+    "engrm": {
+      "url": "http://engrm:3767/mcp",
+      "headers": {
+        "Authorization": "Bearer replace-with-a-long-random-token"
+      }
+    }
+  }
+}
+```
+
+Fleet writes:
+- use the reserved project name `shared-experience` by default
+- stay `shared` unless explicitly overridden
+- sync to the dedicated fleet namespace/key instead of the normal org namespace
+- get an extra outbound scrub pass that redacts hostnames, IPs, and MACs before upload
+
 ---
 
 ## How It Works
@@ -429,6 +477,7 @@ What each tool is good for:
 - `memory_console` gives the quickest project snapshot, including whether continuity is `fresh`, `thin`, or `cold`
 - `resume_thread` is the fastest “get me back into the live thread” path when you want freshness, source, next actions, tool trail, chat, and one exact `load_recall_item(...)` suggestion in one place
 - `resume_thread(agent="claude-code")` lets you deliberately recover one agent's thread on a shared repo instead of only taking the blended repo-level default
+- when Claude Code is active on a shared repo, startup, the recall index, and the default `resume_thread` path now bias toward Claude’s thread first, so Claude sessions are less likely to get nudged into a Codex/OpenClaw branch of the same project by accident
 - startup and the MCP workbench now surface a direct `resume_thread(agent="...")` hint when multiple agents are active on one repo
 - `list_recall_items` is the deterministic directory-first path when you want to inspect the best candidate handoffs/threads before opening one exact item
 - `load_recall_item` completes that protocol by letting agents open one exact recall key directly after listing
@@ -589,6 +638,8 @@ Engrm auto-registers in:
 - `~/.claude/settings.json` — 6 lifecycle hooks
 - `~/.codex/config.toml` — MCP server (`engrm`) + `codex_hooks` feature flag
 - `~/.codex/hooks.json` — SessionStart and Stop hooks
+- `~/.config/opencode/opencode.json` — MCP server (`engrm`)
+- `~/.config/opencode/plugins/engrm.js` — local OpenCode continuity plugin
 
 ---
 
@@ -623,7 +674,7 @@ See [LICENSE](LICENSE) for full terms.
 - Security: [SECURITY.md](SECURITY.md)
 - Roadmap: [ROADMAP.md](ROADMAP.md)
 
-**Maintainers:** run `node scripts/check-public-docs.mjs` to verify the repo only contains the approved public docs set at the root.
+**Maintainers:** run `npm run check:release` before publish to verify the build, public-doc guard, and OpenClaw plugin contract together.
 
 ---