@genclass/mcp 0.1.0 → 0.2.0

package/CHANGELOG.md

@@ -1,6 +1,58 @@
 # Changelog
 
-## 0.1.0 (unreleased)
+## 0.2.0 (2026-10-25)
+
+The `0.1.0` on npm (published 2026-10-25) is an earlier build: eight tools, no `genclass_scan`, no `GENCLASS_REPORT`.
+Everything below ships first in `0.2.0`.
+
+- **`genclass_setup` installs the newest runtime line.** It prints `npm install @genclass/runtime@beta` (the agent
+  entry and interaction analytics need 0.3.0-beta.1 or later) for a project on 0.2.x, on a range without a version
+  (`*`, `latest`: npm's `latest` is still 0.2.x) or with an older copy in `node_modules`, and says what the project
+  has. The README's connect step names `@beta` first.
+- **`genclass_scan {url, phone?, hosted?}`: audit any page in one step** ([docs/runtime/AGENT.md](../../docs/runtime/AGENT.md)
+  section 4.2). It runs `@genclass/runtime`'s `scan --json` (the project's own runtime when it has the command, else
+  `npx -y @genclass/runtime@beta`) in the project's folder: headless Chromium at desktop and phone sizes with the 23
+  measured visual checks, and returns a score and each problem with its element, selector and suggested CSS fix (page
+  text cleaned and labelled as data), plus the report as JSON (a second text block and `structuredContent`). It passes
+  `--local` unless the agent sets `hosted: true`, so without Playwright's Chromium it says how to install it instead
+  of sending the URL anywhere; `hosted: true` uses the genclass.dev page check (public pages only). The server's
+  instructions and the plugin skill say to scan a page before calling a UI change done. Nine tools.
+
+- **Fixed: a race replay counts page lines.** Explore compares page text as counts, so a product name the page
+  always shows once and a stale result showed a second time was "only in the failing run"; the replay verdict then
+  checked presence only, and the card alone made every replay "reproduced", including after the fix. The finding's
+  evidence now keeps the counts (`textCounts`) and the verdict needs the line as many times as the failing run showed
+  it. Found by the end-to-end walkthrough, covered by a test.
+- **Fixed: with `GENCLASS_REPORT` set, the tools' own calls are mirrored too.** The server's in-process API (what
+  `genclass_explain`, `genclass_replay`, `genclass_explore` and `genclass_clear` call when this session owns the
+  bridge) bypassed the reporting wrapper, so a dashboard only ever received the findings; the evidence read, the test,
+  replays armed and their verdicts, explore reports and clears reached it only through the bridge's HTTP path. Found
+  by the end-to-end walkthrough (`docs/evidence/e2e-walkthrough`), covered by a test.
+- **Opt-in reporting to the app's dashboard** (`GENCLASS_REPORT`; [docs/runtime/AGENT.md](../../docs/runtime/AGENT.md)
+  section 4.5). Set to `on` (with `GENCLASS_DASHBOARD`) or to a dashboard link, the server mirrors its log of the loop
+  to that dashboard: each finding the bridge makes, each time the agent reads its evidence (`genclass_explain`), each
+  replay armed and each replay verdict, each regression test written, each explore report (page loads compared, races
+  found) and each clear: `POST /api/projects/<secret>/agent`, batched, best effort, ids, kinds, page paths, verdicts,
+  counts and times only, never page text. The dashboard's Closed-loop development view then draws every issue's path
+  (found, reproduced, fix proven, closed, or round again when a replay after a fix still reproduces it) with the
+  period's counts, lists the log newest first, and shows per production problem what the agent reproduced, tested and
+  proved fixed. Off unless set; nothing sent feeds a decision, a finding or a tool result.
+
+- **Explore stops reporting order-dependent bookkeeping as races** ([docs/runtime/AGENT.md](../../docs/runtime/AGENT.md)
+  section 5.1, [RESULTS.md](../../docs/runtime/RESULTS.md) section 20). A reordered run is a race only when what the
+  page shows, its form values, the requests it sent or its errors differ from the request-order baseline; store values
+  that differ while the page is the same are listed as "Internal state only", and an overwrite a later response
+  repaired as "Along the way only", neither with a finding. A `[redacted]` value equals any value. Scenarios are
+  grouped with digits blanked, parenthesised counts dropped and a typing burst split by a label change joined again,
+  falling back to action kinds and element roles, then kinds, when names changed with the order; the report says how
+  many reordered runs it compared and says "0 compared" (no verdict) rather than "No race found" when none was. Race
+  finding titles, keys, replays and generated tests lead with what the page showed (store values only for a finding
+  with nothing else). Snapshots may carry up to 1,200 store fields and say how many the page left out. The explore
+  tool's description and the plugin skill say the same. On the Strapi admin's 21 interaction sessions the race-like
+  verdicts went from 16 to 0 (reordered runs reported as races: 62, all false, to 0; sessions without a usable verdict:
+  5 to 0), and the real 403 race is still found at 300, 600 and 900 ms (6 of 6 runs, 0 false); on a correct Redux
+  Toolkit Query page the false reports went from 18 of 18 reordered runs to 0; the e2e page still finds its race in 5
+  of 6 reordered runs and none on the fixed page.
 
 - First version: the GenClass MCP server for coding agents ([docs/runtime/AGENT.md](../../docs/runtime/AGENT.md)).
   A bridge on 127.0.0.1:47357 that pages connect to through `@genclass/runtime/agent`; eight tools
@@ -15,3 +67,17 @@
   agent and strips invisible characters. A Claude Code plugin (`plugin/`: the server, the prompt hook and a skill) and
   an MCP Registry entry (`server.json`) ship with it. Needs `@genclass/runtime` 0.3.0-beta.1 or later in the app
   (`@genclass/runtime/agent`). Released by `.github/workflows/release-mcp.yml` from a `mcp-v<version>` tag.
+- Ready for registries and directories: a `server.json` description of at most 100 characters (the official MCP
+  Registry's limit; checked against its 2025-12-11 schema and accepted by `mcp-publisher validate`), a `title` and
+  `websiteUrl`; the release workflow publishes it to the registry with GitHub OIDC (pinned, checksummed
+  `mcp-publisher`) after npm, and uploads an MCP Bundle (`npm run bundle`: `genclass-<version>.mcpb`, for Claude
+  Desktop and Smithery) as an artifact. Every launcher is pinned to the release (`npx -y @genclass/mcp@<version>`):
+  the plugin's server and hook, `init`, and the install blocks; `npm run sync` rewrites the pins from
+  `package.json` and the tests fail when one differs. Install blocks for Claude Code, Claude Desktop, Cursor (with an
+  install link), VS Code (`code --add-mcp` and an install link), Codex CLI, Gemini CLI, Windsurf, Cline, Goose, Zed,
+  JetBrains IDEs and OpenCode, printed by `genclass-mcp config [client]` and in the README; `genclass_setup` adds the
+  block for the client it runs in (from `clientInfo` at initialize). Every tool's annotations carry `title`,
+  `readOnlyHint`, `destructiveHint` and `openWorldHint`. The plugin has a README (what it installs, what the hook
+  sends where, how to turn it off), directory links in `plugin.json`, and an [Agent Plugins](https://agent-plugins.org)
+  `plugin.json` and `mcp.json` for Cursor and Codex; the repository root has `glama.json`, `gemini-extension.json`,
+  `llms-install.md` and `.cursor-plugin/marketplace.json`. Listing drafts: `docs/adoption/listings/mcp.md`.

package/README.md

@@ -15,28 +15,261 @@ An MCP server for Claude Code and other coding agents. It connects to the pages
 - **Proof.** `genclass_replay` delivers the exact order that produced a bug again, before your fix (reproduced) and
   after it (not reproduced), and `genclass_explain` writes a Playwright test that fails while the bug is there.
 
-Everything stays on your machine. No dependencies.
+Everything stays on your machine. No dependencies. The server is free. For developers who want the loop run for
+them, GenClass runs it as a service under Enterprise: the MCP server, collector and dashboard hosted by GenClass or
+inside your network ([genclass.dev/enterprise](https://genclass.dev/enterprise)).
 
 ## Install
 
-As a Claude Code plugin (the server, a prompt hook that brings new findings to each prompt, and a skill that tells
-Claude when to use it):
+Pick your client. Every block pins this release, so the client starts exactly this version;
+`npx -y @genclass/mcp@0.2.0 config <client>` prints the same lines, and `genclass_setup` prints the block for the
+client it runs in.
+
+<!-- install:start (generated from src/clients.ts by `npm run sync -w @genclass/mcp`; edit it there) -->
+
+### Claude Code
+
+As a plugin (the server, a prompt hook that brings new findings to each prompt, and a skill that tells Claude when to use it):
 
 ```
 /plugin marketplace add genclass-dev/GenClass-lib
 /plugin install genclass@genclass
 ```
 
-Or the server alone:
+Or the server alone (`--scope project` writes `.mcp.json` for everyone on the project; `--scope user` for all your projects):
+
+```bash
+claude mcp add --scope project genclass -- npx -y @genclass/mcp@0.2.0
+# or: npx -y @genclass/mcp@0.2.0 init --hook   (.mcp.json, plus the prompt hook in .claude/settings.json)
+```
+
+### Claude Desktop
+
+Settings, Developer, Edit Config (`~/Library/Application Support/Claude/claude_desktop_config.json` on macOS, `%APPDATA%\Claude\claude_desktop_config.json` on Windows):
+
+```json
+{
+  "mcpServers": {
+    "genclass": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "@genclass/mcp@0.2.0"
+      ]
+    }
+  }
+}
+```
+
+### Cursor
+
+[![Add genclass MCP server to Cursor](https://cursor.com/deeplink/mcp-install-dark.png)](https://cursor.com/link/mcp/install?name=genclass&config=eyJjb21tYW5kIjoibnB4IiwiYXJncyI6WyIteSIsIkBnZW5jbGFzcy9tY3BAMC4yLjAiXX0%3D) or `cursor://anysphere.cursor-deeplink/mcp/install?name=genclass&config=eyJjb21tYW5kIjoibnB4IiwiYXJncyI6WyIteSIsIkBnZW5jbGFzcy9tY3BAMC4yLjAiXX0%3D`
+
+Or `.cursor/mcp.json` in the project (`~/.cursor/mcp.json` for all projects):
+
+```json
+{
+  "mcpServers": {
+    "genclass": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "@genclass/mcp@0.2.0"
+      ]
+    }
+  }
+}
+```
+
+### VS Code
+
+[![Install in VS Code](https://img.shields.io/badge/VS_Code-VS_Code?style=flat-square&label=Install%20Server&color=0098FF)](https://vscode.dev/redirect/mcp/install?name=genclass&config=%7B%22command%22%3A%22npx%22%2C%22args%22%3A%5B%22-y%22%2C%22%40genclass%2Fmcp%400.2.0%22%5D%7D) or from a terminal:
 
 ```bash
-claude mcp add genclass -- npx -y @genclass/mcp
-# or, in the project: npx @genclass/mcp init          (adds it to .mcp.json)
-#                     npx @genclass/mcp init --hook   (also a prompt hook that brings new findings to the agent)
+code --add-mcp '{"name":"genclass","command":"npx","args":["-y","@genclass/mcp@0.2.0"]}'
 ```
 
-Then connect the app's development build (once). The agent link needs `@genclass/runtime` 0.3.0-beta.1 or later
-(`npm install @genclass/runtime@beta` until 0.3.0 is `latest`):
+Or `.mcp.json` at the workspace root (the portable format; `.vscode/mcp.json` takes the same entry under `servers`):
+
+```json
+{
+  "mcpServers": {
+    "genclass": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "@genclass/mcp@0.2.0"
+      ]
+    }
+  }
+}
+```
+
+### Codex CLI
+
+```bash
+codex mcp add genclass -- npx -y @genclass/mcp@0.2.0
+```
+
+Or `~/.codex/config.toml` (`.codex/config.toml` in a trusted project). The first start downloads the package, so allow more than the default 10 seconds:
+
+```toml
+[mcp_servers.genclass]
+command = "npx"
+args = ["-y", "@genclass/mcp@0.2.0"]
+startup_timeout_sec = 30
+```
+
+### Gemini CLI
+
+```bash
+gemini mcp add genclass npx -- -y @genclass/mcp@0.2.0   # -s user: for all projects
+```
+
+Or `.gemini/settings.json` in the project (`~/.gemini/settings.json` for all projects):
+
+```json
+{
+  "mcpServers": {
+    "genclass": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "@genclass/mcp@0.2.0"
+      ]
+    }
+  }
+}
+```
+
+### Windsurf (Devin Desktop)
+
+`~/.codeium/windsurf/mcp_config.json` (Windsurf), `~/.config/devin/mcp_config.json` (Devin Desktop's Cascade agent; `%APPDATA%\devin\mcp_config.json` on Windows):
+
+```json
+{
+  "mcpServers": {
+    "genclass": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "@genclass/mcp@0.2.0"
+      ]
+    }
+  }
+}
+```
+
+Devin CLI and Devin Desktop's Devin Local agent:
+
+```bash
+devin mcp add -s user genclass -- npx -y @genclass/mcp@0.2.0
+```
+
+### Cline
+
+```bash
+cline mcp add genclass --yes -- npx -y @genclass/mcp@0.2.0
+```
+
+Or MCP Servers, Configure, Configure MCP Servers (`~/.cline/data/settings/cline_mcp_settings.json`):
+
+```json
+{
+  "mcpServers": {
+    "genclass": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "@genclass/mcp@0.2.0"
+      ]
+    }
+  }
+}
+```
+
+### Goose
+
+`goose://extension?cmd=npx&arg=-y&arg=%40genclass%2Fmcp%400.2.0&id=genclass&name=genclass&description=Find%20race%20conditions%20by%20reordering%20responses%2C%20replay%20one%2C%20prove%20the%20fix` (open it to install), or `goose configure` (Add Extension, Command-line Extension, command `npx -y @genclass/mcp@0.2.0`), or `~/.config/goose/config.yaml`:
+
+```yaml
+extensions:
+  genclass:
+    type: stdio
+    name: genclass
+    enabled: true
+    cmd: npx
+    args: ["-y", "@genclass/mcp@0.2.0"]
+    env_keys: []
+    envs: {}
+    timeout: 300
+```
+
+### Zed
+
+`~/.config/zed/settings.json` (`.zed/settings.json` in the project), or Settings, AI, MCP Servers, Add Server:
+
+```json
+{
+  "context_servers": {
+    "genclass": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "@genclass/mcp@0.2.0"
+      ]
+    }
+  }
+}
+```
+
+### JetBrains IDEs
+
+AI Assistant: Settings, Tools, AI Assistant, Model Context Protocol (MCP), Add, then this JSON. Junie: `.junie/mcp/mcp.json` in the project (`~/.junie/mcp/mcp.json` for all projects), same JSON:
+
+```json
+{
+  "mcpServers": {
+    "genclass": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "@genclass/mcp@0.2.0"
+      ]
+    }
+  }
+}
+```
+
+### OpenCode
+
+`opencode.json` in the project (`~/.config/opencode/opencode.json` for all projects), or `opencode mcp add`:
+
+```json
+{
+  "$schema": "https://opencode.ai/config.json",
+  "mcp": {
+    "genclass": {
+      "type": "local",
+      "command": [
+        "npx",
+        "-y",
+        "@genclass/mcp@0.2.0"
+      ],
+      "enabled": true
+    }
+  }
+}
+```
+
+<!-- install:end -->
+
+Any other MCP client: a stdio server, command `npx`, arguments `-y @genclass/mcp@0.2.0`, no environment variables
+required.
+
+Then connect the app's development build (once). The agent link needs the newest runtime line, `@genclass/runtime`
+0.3.0-beta.1 or later: `npm install @genclass/runtime@beta` (a bare `npm install @genclass/runtime` gets npm's
+`latest`, the older 0.2.x, which has no agent entry):
 
 ```ts
 import { GenClass } from "@genclass/runtime";
@@ -76,7 +309,8 @@ on the fixed one, and explore on the fixed page reported nothing. Details: [docs
 | `genclass_explore` | `start` / `report` / `stop`: find what depends on response order |
 | `genclass_replay` | `arm` / `result`: reproduce one finding exactly |
 | `genclass_analytics` | per-endpoint latency, error rates and request counts, learned relations and flows; production numbers from your GenClass dashboard |
-| `genclass_setup` | the lines to add for your framework |
+| `genclass_setup` | the lines to add for your framework, and the config lines for the client it runs in |
+| `genclass_scan` | `{url, phone?, hosted?}`: audit any page (the dev server or a site you do not own) with the 23 measured visual checks in headless Chromium: a score and each problem with its element, selector and a suggested CSS fix, plus the JSON report. Runs `npx @genclass/runtime@beta scan --json --local` (or the project's own runtime); needs Playwright's Chromium (`npx playwright install chromium`), or `hosted: true` for the genclass.dev page check (public pages only) |
 | `genclass_clear` | forget recorded runs and findings |
 
 ## New findings without asking
@@ -84,8 +318,8 @@ on the fixed one, and explore on the fixed page reported nothing. Details: [docs
 - **Channels** (Claude Code research preview): start Claude Code with
   `claude --dangerously-load-development-channels server:genclass` and new findings arrive in the session as they
   happen (`notifications/claude/channel`).
-- **Prompt hook**: `npx @genclass/mcp init --hook` adds a `UserPromptSubmit` hook; each prompt then carries the findings
-  that changed since the last one.
+- **Prompt hook** (part of the Claude Code plugin): `npx -y @genclass/mcp@0.2.0 init --hook` adds a `UserPromptSubmit`
+  hook to `.claude/settings.json`; each prompt then carries the findings that changed since the last one.
 
 ## Environment
 
@@ -94,6 +328,7 @@ on the fixed one, and explore on the fixed page reported nothing. Details: [docs
 | `GENCLASS_AGENT_PORT` | `47357` | the bridge's port (the page then needs `connectAgent(rt, { url })`) |
 | `GENCLASS_STATE_DIR` | `.genclass/agent` | where findings persist between sessions; `none` keeps them in memory |
 | `GENCLASS_DASHBOARD` | | your dashboard's private link, for `genclass_analytics` `production` (`sample` reads the public sample) |
+| `GENCLASS_REPORT` | | `on` (with `GENCLASS_DASHBOARD`), or a dashboard link or secret: mirror this server's log of the loop to that dashboard's Closed-loop development view (each finding, each time its evidence is read, each replay armed and its verdict, each regression test written, each explore report, each clear), which draws every issue's path (found, reproduced, fix proven, closed, or round again) with counts and lists the log. Ids, kinds, page paths, verdicts, counts and times only, never page text; off unless set ([AGENT.md section 4.5](https://github.com/genclass-dev/GenClass-lib/blob/main/docs/runtime/AGENT.md)) |
 | `GENCLASS_AGENT_ORIGINS` | | exact page origins allowed besides this machine's, e.g. `http://192.168.1.20:5173` (the page then needs `connectAgent(rt, { anyHost: true })`) |
 | `GENCLASS_TOKEN_DIR` | `~/.genclass` | where the bridge keeps its API token (mode 0600) for the hook and other sessions |
 
@@ -104,9 +339,26 @@ The bridge listens on 127.0.0.1 only and checks the `Host` header. Pages may tal
 website, a router page or a captive portal you visit cannot plant findings, and a page can only add to its own run.
 The agent API refuses browsers and needs a token only your user can read. Everything a page sends is validated,
 capped, stripped of invisible characters and shown to the agent as quoted data, never as instructions. Nothing leaves
-your machine. See section 6 of
+your machine (the exceptions are under Privacy). See section 6 of
 [AGENT.md](https://github.com/genclass-dev/GenClass-lib/blob/main/docs/runtime/AGENT.md).
 
+## Privacy
+
+The server collects nothing and sends nothing to GenClass or anyone else. Findings stay on your machine, in
+`.genclass/agent/` in the server's working directory (the project, for clients that start servers there; set
+`GENCLASS_STATE_DIR` for clients that do not, such as Claude Desktop), which the server gitignores, and in the bridge's memory; the bridge's token is in
+`~/.genclass`. The one request that leaves the machine is `genclass_analytics` `{source: "production"}`, which reads
+your own GenClass dashboard when you set `GENCLASS_DASHBOARD` to its private link. `genclass_scan` loads the page the
+agent names (that is the page's own traffic) in a local browser; only with `hosted: true` is the URL sent to the
+genclass.dev page check, and when the project has no `@genclass/runtime` with `scan`, npx downloads it from npm. With `GENCLASS_REPORT` set (off by
+default) the server also mirrors its log of the loop to that dashboard, in small batches: the id, kind and page path of
+each finding, when its evidence was read, each replay armed and its verdict, when a regression test was written, each
+explore report (page loads compared, races found) and each clear, so the dashboard's Closed-loop development view
+draws every issue's path (found, reproduced, fix proven, closed, or round again) and lists what the agent did; never a
+title, an evidence line, a snapshot or any page text, and nothing it sends feeds a decision. The runtime in your app has its
+own disclosed diagnostics ([PRIVACY.md](https://github.com/genclass-dev/GenClass-lib/blob/main/PRIVACY.md)), which
+`GenClass.init({ telemetry: false })` turns off; the agent link is not part of them.
+
 ## License
 
 Apache-2.0

package/dist/cli.d.ts

@@ -10,10 +10,12 @@ declare function init(io: CliIo, hook: boolean): number;
  * epoch (a restarted bridge counts from 0 again). Silent when there is nothing new or no bridge.
  */
 declare function hookCmd(io: CliIo): Promise<number>;
+/** `config [client]`: one client's install block, or every client's. */
+declare function configCmd(io: CliIo, id: string | undefined): number;
 declare function runCli(argv: string[], io: CliIo & {
     stdin: NodeJS.ReadableStream;
     stdout: NodeJS.WritableStream;
     stderr: NodeJS.WritableStream;
 }): Promise<number>;
 
-export { type CliIo, hookCmd, init, runCli };
+export { type CliIo, configCmd, hookCmd, init, runCli };