docs: add CHANGELOG.md documenting breaking changes in v0.24.0 (bd-pdr2)

Created CHANGELOG.md for beads-mcp documenting:
- Breaking change: ready() and list() now return list[IssueMinimal] | CompactedResult
- Migration guide for clients expecting list[Issue]
- Context engineering optimization details (~80% context reduction)

Assessment complete: Current approach is correct:
- Type "mismatch" is intentional two-layer architecture
- Client layer returns List[Issue], MCP layer converts to minimal
- Adding full=True would defeat context optimization purpose
- CONTEXT_ENGINEERING.md already has migration guide

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

integrations/beads-mcp/CHANGELOG.md

@@ -0,0 +1,88 @@
+# Changelog
+
+All notable changes to beads-mcp will be documented in this file.
+
+The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
+and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
+
+## [0.41.0] - 2024-12-30
+
+### Changed
+- Minor maintenance release
+
+## [0.24.0] - 2024-01-15
+
+### Changed
+
+#### Breaking: Context Engineering Optimizations
+
+The `ready()` and `list()` MCP tools now return `list[IssueMinimal]` or `CompactedResult` instead of `list[Issue]`. This **breaking change** reduces context window usage by ~80%.
+
+**What changed:**
+- `ready()` returns `list[IssueMinimal] | CompactedResult` (was `list[Issue]`)
+- `list()` returns `list[IssueMinimal] | CompactedResult` (was `list[Issue]`)
+- `show()` still returns full `Issue` (unchanged)
+
+**IssueMinimal includes:**
+- `id`, `title`, `status`, `priority`, `issue_type`
+- `assignee`, `labels`, `dependency_count`, `dependent_count`
+
+**IssueMinimal excludes (use show() for these):**
+- `description`, `design`, `acceptance_criteria`, `notes`
+- `created_at`, `updated_at`, `closed_at`
+- `dependencies`, `dependents` (full objects)
+
+**CompactedResult (returned when >20 results):**
+```python
+{
+    "compacted": True,
+    "total_count": 47,
+    "preview": [/* first 5 IssueMinimal */],
+    "preview_count": 5,
+    "hint": "Use show(issue_id) for full details"
+}
+```
+
+**Migration guide:**
+
+1. Check for compacted results:
+   ```python
+   if isinstance(response, dict) and response.get("compacted"):
+       issues = response["preview"]
+   else:
+       issues = response
+   ```
+
+2. Use `show()` for full details:
+   ```python
+   for issue in issues:
+       full_issue = show(issue_id=issue.id)
+       print(full_issue.description)
+   ```
+
+3. Available options for list operations:
+   - `brief=True` - Returns `BriefIssue` (id, title, status, priority only)
+   - `fields=["id", "title"]` - Custom field projection
+   - `max_description_length=100` - Truncate descriptions
+
+**Rationale:**
+- Reduces context usage from ~400 bytes/issue to ~80 bytes/issue
+- Prevents context overflow for large issue lists (>20 items)
+- Encourages efficient "list then show" pattern for AI agents
+
+See [CONTEXT_ENGINEERING.md](./CONTEXT_ENGINEERING.md) for full migration guide.
+
+### Added
+- `discover_tools()` - Lightweight tool catalog for lazy schema loading
+- `get_tool_info(tool_name)` - On-demand tool details
+- `CompactedResult` model for large result sets
+- `IssueMinimal` model for list views
+- `BriefIssue` model for ultra-compact responses
+- `brief`, `fields`, `max_description_length` parameters for all list operations
+- Environment variables for compaction tuning:
+  - `BEADS_MCP_COMPACTION_THRESHOLD` (default: 20)
+  - `BEADS_MCP_PREVIEW_COUNT` (default: 5)
+
+## [0.23.0] and earlier
+
+See git history for changes prior to context engineering optimizations.