@cntxt-labs/medha-cli 0.7.3 → 0.7.4

package/README.md

@@ -228,6 +228,12 @@ Register the server in your project's `.mcp.json`:
 { "mcpServers": { "medha": { "command": "medha", "args": ["mcp", "serve"] } } }
 ```
 
+Or let medha write it for your client: `medha mcp config --list` shows the clients it knows
+(Claude Code, Cursor, GitHub Copilot / VS Code, opencode, Claude Desktop) with their scopes, and
+`medha mcp config <client> --write` merges the entry into that client's file without touching the
+other servers there. Without `--write` it prints the snippet; `--scope user`, `--launcher npx` and
+`--all` change what it renders.
+
 Then copy [`SKILL.md`](SKILL.md) to
 `.claude/skills/medha/SKILL.md` so the agent knows when and how to use it.
 
@@ -244,6 +250,17 @@ Then copy [`SKILL.md`](SKILL.md) to
 | `simulate` | Preview a signal's effect; persists nothing. |
 | `status` | Engine health and preflight. |
 
+### Claude Code plugin
+
+In Claude Code, the plugin brings the MCP server (pinned to this version, through `npx`), the agent
+skill and four read-side slash commands (`/medha:hints`, `/medha:status`, `/medha:drift`,
+`/medha:pack`) in one install:
+
+```text
+/plugin marketplace add nimishph/cntxt-labs
+/plugin install medha@cntxt-labs
+```
+
 ### Fitting trust into a prompt
 
 `medha pack --budget 2000` selects the active and probation rules that fit a token budget, so a
@@ -324,7 +341,9 @@ medha maintain restore snapshot.json
 ```
 
 Everything lives under `.medha/`. `medha init` scaffolds a `.medha/.gitignore` and a
-`.medha/README.md` for you: `config.json` (the registries) is meant to be committed, while the
+`.medha/README.md` for you, and keeps a short medha section in the project's `AGENTS.md` /
+`CLAUDE.md` (or `--agents-file <path>`; `--no-agents-file` for none) so coding agents know to use
+it. Re-run `medha init` after an upgrade: the store is left alone and that section is refreshed. `config.json` (the registries) is meant to be committed, while the
 store data itself is gitignored by default — share it with `medha sync` instead, since it has its
 own conflict-resolving merge, not git's. Delete or edit `.medha/.gitignore` if you'd rather commit
 the store file directly as a simpler, manual sync.

package/SKILL.md

@@ -47,7 +47,10 @@ You record evidence; **you** decide what to do with it.
 
 ## Setup
 
-Run once per project (creates `.medha/`; add it to `.gitignore` or commit it deliberately):
+Run once per project (creates `.medha/`; add it to `.gitignore` or commit it deliberately). It also
+keeps a short medha section in the project's `AGENTS.md` / `CLAUDE.md` (`--agents-file <path>` to
+name another file, `--no-agents-file` for none). Running it again after an upgrade is safe: the store
+is left alone and the section is refreshed to the new version.
 
 ```sh
 medha init

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@cntxt-labs/medha-cli",
-  "version": "0.7.3",
+  "version": "0.7.4",
   "description": "Evidential memory for rules, recipes and tools: records what happened and returns trust hints. CLI and MCP server.",
   "keywords": [
     "mcp",
@@ -40,11 +40,11 @@
     "node": ">=18"
   },
   "optionalDependencies": {
-    "@cntxt-labs/medha-linux-x64": "0.7.3",
-    "@cntxt-labs/medha-linux-arm64": "0.7.3",
-    "@cntxt-labs/medha-darwin-arm64": "0.7.3",
-    "@cntxt-labs/medha-darwin-x64": "0.7.3",
-    "@cntxt-labs/medha-win32-x64": "0.7.3"
+    "@cntxt-labs/medha-linux-x64": "0.7.4",
+    "@cntxt-labs/medha-linux-arm64": "0.7.4",
+    "@cntxt-labs/medha-darwin-arm64": "0.7.4",
+    "@cntxt-labs/medha-darwin-x64": "0.7.4",
+    "@cntxt-labs/medha-win32-x64": "0.7.4"
   },
   "devDependencies": {
     "@modelcontextprotocol/sdk": "^1.30.0",