Add report_issue tests and update tool counts in docs

- Add 45 tests covering input coercion, JSONL safety (special chars,
  injection attempts, tracebacks), file I/O, rotation, and validation
- Update tool count from 53/54 to 55 in CLAUDE.md, README.md, and docs
- Add report_issue to CLAUDE.md tool table
- Add report_issue section to collaboration docs page

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: ianhi

CLAUDE.md

@@ -15,7 +15,7 @@ A TypeScript MCP server that connects to JupyterLab's real-time collaboration sy
 ```
 src/
 ├── index.ts        # Thin MCP server entry point (dispatches to handlers)
-├── handlers/       # All 54 tool handlers, organized by category
+├── handlers/       # All 55 tool handlers, organized by category
 │   ├── connection.ts   # connect_jupyter, list_files, list_notebooks, list_kernels, open/create/rename_notebook
 │   ├── cell-read.ts    # get_notebook_content, get_notebook_outline, search_notebook, replace_in_notebook
 │   ├── cell-write.ts   # insert_cell, update_cell, delete_cell(s), change_cell_type, copy/move_cells, batch ops
@@ -24,7 +24,7 @@ src/
 │   ├── kernel-lsp.ts   # kernel status/variables/interrupt/restart, diagnostics, hover_info, rename_symbol, diff/rename_notebook
 │   └── collab.ts       # get_user_focus, cell history, notebook changes, recover_cell, snapshots, locks
 ├── connection.ts   # JupyterLab connection state, config, session management, kernel execution
-├── schemas.ts      # Tool schema definitions (all 53 tools)
+├── schemas.ts      # Tool schema definitions (all 55 tools)
 ├── tool-helpers.ts # Shared handler patterns (getNotebookCells, resolveIndexParam, etc.)
 ├── helpers.ts      # Shared utilities (cell extraction, diffing, output formatting)
 ├── notebook-fs.ts  # Filesystem backend (read/write .ipynb without JupyterLab)
@@ -93,6 +93,7 @@ src/
 | `lock_cells` | Acquire advisory locks on cells |
 | `unlock_cells` | Release advisory locks |
 | `list_locks` | List active cell locks |
+| `report_issue` | Submit a feedback report (persisted to JSONL) |
 
 ### Cell IDs (Stable Addressing)
 

README.md

@@ -11,7 +11,7 @@ Claude Code  ←—stdio—→  MCP Server  ←—y-websocket—→  JupyterLab
 Claude Code can already edit files, but notebooks are special — they have cells, kernels, outputs, and a live browser UI. This MCP server bridges the gap:
 
 - **Real-time sync** — edits appear instantly in JupyterLab via y-websocket
-- **53 tools** — read, edit, execute, search, diff, tag, lock, snapshot, and more
+- **55 tools** — read, edit, execute, search, diff, tag, lock, snapshot, and more
 - **Cell ID addressing** — stable references that survive insertions and deletions
 - **Multi-agent ready** — cell locking, change tracking, and per-agent attribution
 - **Context-efficient** — filter by cell type, skip outputs, limit images
@@ -73,6 +73,52 @@ The docs site has detailed tool reference pages, parameter tables, examples, and
 
 See also [datalayer/jupyter-mcp-server](https://github.com/datalayer/jupyter-mcp-server) — a Python-based alternative with Streamable HTTP transport and streaming execution. See the [comparison page](https://ianhi.github.io/jupyterlab-collab-mcp/comparison/) for detailed differences.
 
+## Development
+
+### Setup
+
+```bash
+git clone https://github.com/ianhi/jupyterlab-collab-mcp.git
+cd jupyterlab-collab-mcp
+npm install
+```
+
+### Dev workflow
+
+Point the MCP server at your local build and run the TypeScript compiler in watch mode:
+
+```bash
+claude mcp add -s user jupyter -- node $PWD/dist/index.js
+npm run build    # initial build
+npm run watch    # recompile on file changes (run in a separate terminal)
+```
+
+After each recompile, run `/mcp` in Claude Code to reconnect to the updated server. No need to remove and re-add the MCP config.
+
+> **Why not `tsx --watch`?** MCP servers are long-running stdio processes. `tsx --watch` detects the server as "completed" and kills it, breaking the connection.
+
+### Build + test
+
+```bash
+npm run build          # compile TypeScript to dist/
+npm run watch          # same, but recompile on changes
+npm test               # run unit tests (vitest)
+npm run test:watch     # run tests in watch mode
+```
+
+### Integration testing
+
+Requires a running JupyterLab with `jupyter-collaboration`:
+
+```bash
+jlabx                                    # start JupyterLab (or jupyter lab)
+JUPYTER_TOKEN=<token> npm run test:integration   # run src/test.ts against it
+```
+
+### After changing tool schemas
+
+If you add/remove tools or change their parameters, run `/mcp` in Claude Code to reconnect. The server caches tool definitions at startup, so a reconnect is needed for schema changes to take effect.
+
 ## License
 
 BSD-3-Clause

docs/src/content/docs/comparison.md

@@ -59,7 +59,7 @@ get_user_focus(path) → { focusedCell: 3, cursorPosition: 42 }
 
 Cell locking, change tracking, named snapshots, and per-agent attribution make it safe to run multiple agents on the same notebook. See the [Multi-Agent Guide](../agents/multi-agent/).
 
-### 53 tools
+### 55 tools
 
 Comprehensive coverage: batch operations, cross-notebook copy/move, cell ID addressing, metadata/tags, diagnostics, symbol rename, and more.
 

docs/src/content/docs/tools/collaboration.md

@@ -185,4 +185,25 @@ Shows which cells were added, deleted, modified, or unchanged. Modified cells in
 
 ---
 
+## Issue reporting
+
+### report_issue
+
+Submit a feedback report about a tool bug, hang, missing feature, or general observation. Reports are persisted to a JSONL file (`~/.jupyter-mcp-reports.jsonl`) for developer review.
+
+| Parameter | Type | Required | Description |
+|-----------|------|----------|-------------|
+| `category` | string | Yes | One of: `tool_bug`, `hang`, `missing_feature`, `observation`, `user_feedback` |
+| `summary` | string | Yes | One-line description |
+| `tool_name` | string | No | Which MCP tool was involved |
+| `path` | string | No | Notebook path |
+| `details` | string | No | Error messages or reproduction steps |
+
+**Notes:**
+- All inputs are defensively coerced to strings and truncated (summary: 500 chars, details: 2000 chars) to keep writes atomic
+- The reports file is automatically rotated when it exceeds 1MB
+- Safe for concurrent writes from multiple agents (uses `O_APPEND`)
+
+---
+
 **See also:** [Metadata & tags](../metadata/), [Kernel & analysis tools](../kernel/)

docs/src/content/docs/tools/index.md

@@ -1,9 +1,9 @@
 ---
 title: Tool Overview
-description: All 54 MCP tools organized by category.
+description: All 55 MCP tools organized by category.
 ---
 
-The MCP server provides 54 tools for working with Jupyter notebooks. Tools are organized into categories below — click through for full parameter documentation and examples.
+The MCP server provides 55 tools for working with Jupyter notebooks. Tools are organized into categories below — click through for full parameter documentation and examples.
 
 ## Categories