fix: add MCP server config to ExpandChatSession for feature creation

[sundog75]

Problem

When adding features to existing projects via the "Add Features" / expand project flow, Claude reports it "could not access the MCP tool required to add features" and asks users to add them manually.

Root cause: The ExpandChatSession class does NOT configure an MCP server, while AssistantChatSession does. This means Claude cannot use feature_create_bulk or feature_create MCP tools during project expansion.

Solution

Added MCP server configuration to ExpandChatSession, matching the pattern used in AssistantChatSession:

• Added EXPAND_FEATURE_TOOLS constant for feature_create, feature_create_bulk, and feature_get_stats
• Added MCP server configuration pointing to mcp_server.feature_mcp
• Updated allowed_tools and security settings to include feature tools

Testing

1. Start the UI
2. Select an existing project with features
3. Click "Add Features" or use the expand feature flow
4. Ask Claude to add a new feature
5. Verify feature appears in the Kanban board (no manual addition required)

Notes

• The existing XML parsing fallback (regex matching <features_to_create> tags) continues to work as defense-in-depth if MCP fails
• Relates to Clarification needed: allowed tools for "add feature with AI" dialog #86 (permissions issues in expand feature flow)

Summary by CodeRabbit

• New Features

  • Chat sessions now include additional feature tools at session start.
  • Multi-processor server support enabled to power those tools.
  • Added a tool to fetch passing features for regression workflows, with automatic regression count tracking.

• Chores

  • Added legacy-compatible database fields and a migration to preserve data and track regression counts.

_{✏️ Tip: You can customize this high-level summary in your review settings.}

[coderabbitai]

📝 Walkthrough

Walkthrough

Added EXPAND_FEATURE_TOOLS and MCP setup to the expand chat session service; initialized ClaudeSDKClient with mcp_servers and updated allowed tools; added Feature.regression_count plus legacy columns and migration support; added MCP tool feature_get_for_regression(limit) for selecting and incrementing passing features.

Changes

Cohort / File(s)	Summary
Expand chat session service server/services/expand_chat_session.py	Added EXPAND_FEATURE_TOOLS constant; imported sys; built mcp_servers entry to run mcp_server.feature_mcp with PROJECT_DIR/PYTHONPATH env; passed mcp_servers into ClaudeSDKClient; included EXPAND_FEATURE_TOOLS in security allowed_tools and initial allowed tools.
Database — Feature model & migration api/database.py	Added columns to Feature: regression_count: Column(Integer, nullable=False, default=0, index=True), plus legacy testing_in_progress: Column(Boolean, nullable=False, default=False) and last_tested_at: Column(DateTime, nullable=True, default=None); updated to_dict() to include regression_count; added migration helper to add regression_count to existing DBs and invoked it during DB init/create.
MCP server — regression tool mcp_server/feature_mcp.py	Added public tool feature_get_for_regression(limit: Annotated[int, Field(default=3, ge=1, le=10)] = 3) -> str that selects passing features ordered by regression_count, acquires row locks, increments regression_count, commits/refreshes, and returns a JSON payload; removed RegressionInput model.

Sequence Diagram(s)

sequenceDiagram
    participant Client
    participant ExpandService as "Expand Service"
    participant Env as Environment
    participant ClaudeClient as "ClaudeSDKClient"
    participant MCP as "feature_mcp (MCP Server)"
    participant DB as Database

    Client->>ExpandService: request expand chat session
    ExpandService->>Env: prepare env vars (PROJECT_DIR, PYTHONPATH)
    ExpandService->>MCP: declare/start feature_mcp via mcp_servers
    ExpandService->>ClaudeClient: init with mcp_servers + EXPAND_FEATURE_TOOLS
    ClaudeClient->>MCP: invoke feature tool (feature_get_for_regression)
    MCP->>DB: query passing features ordered by regression_count
    DB-->>MCP: return features
    MCP->>DB: increment regression_count, commit
    DB-->>MCP: confirm update
    MCP-->>ClaudeClient: return feature payload
    ClaudeClient-->>ExpandService: receive tool result
    ExpandService-->>Client: return expanded session

Loading

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Poem

🐰 I hopped through code with nimble feet,

counting regressions, tools to meet,
MCP whispers, DB replies,
Claude and I trade bright surprise,
a rabbit cheers for every feat.

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately reflects the main changes: adding MCP server configuration to ExpandChatSession to enable feature creation functionality.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

• 📝 Generate docstrings

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Fix all issues with AI agents

In `@api/database.py`:
- Around line 61-66: The migration function _migrate_add_testing_columns is
currently a no-op but must add the deprecated columns for DBs created after
their removal; update _migrate_add_testing_columns to check the target table
schema (e.g., via inspector or PRAGMA) for the absence of testing_in_progress
and last_tested_at and, if missing, execute DDL to ALTER TABLE and add
testing_in_progress BOOLEAN NOT NULL DEFAULT FALSE and last_tested_at DATETIME
NULL (or equivalent for SQLite), ensuring the column definitions match the
SQLAlchemy model defaults and nullability and that the operation is safe (wrap
in try/except and commit/rollback as in existing migration patterns). Ensure the
function references the same table name used by the model and handles both
SQLite and other RDBMS dialect nuances when issuing ALTER statements.

[coderabbitai]

⚠️ Potential issue | 🟠 Major

Migration gap: databases created after column removal won't have these columns.

The columns are added to the model, but _migrate_add_testing_columns() (lines 241-253) is a no-op. This creates a problem for databases that were created after the columns were removed in commit 486979c but before this PR—they won't have these columns, and INSERT operations will fail.

Consider updating _migrate_add_testing_columns to actually add the columns if missing:

Proposed fix for _migrate_add_testing_columns

 def _migrate_add_testing_columns(engine) -> None:
-    """Legacy migration - no longer adds testing columns.
-
-    The testing_in_progress and last_tested_at columns were removed from the
-    Feature model as part of simplifying the testing agent architecture.
-    Multiple testing agents can now test the same feature concurrently
-    without coordination.
-
-    This function is kept for backwards compatibility but does nothing.
-    Existing databases with these columns will continue to work - the columns
-    are simply ignored.
-    """
-    pass
+    """Add testing columns for backward compatibility.
+
+    These columns are deprecated but required in the model so SQLAlchemy
+    populates them on INSERT (preventing NOT NULL violations in old databases).
+    """
+    with engine.connect() as conn:
+        result = conn.execute(text("PRAGMA table_info(features)"))
+        columns = [row[1] for row in result.fetchall()]
+
+        if "testing_in_progress" not in columns:
+            conn.execute(text(
+                "ALTER TABLE features ADD COLUMN testing_in_progress BOOLEAN DEFAULT 0"
+            ))
+        if "last_tested_at" not in columns:
+            conn.execute(text(
+                "ALTER TABLE features ADD COLUMN last_tested_at DATETIME DEFAULT NULL"
+            ))
+        conn.commit()

🤖 Prompt for AI Agents

In `@api/database.py` around lines 61 - 66, The migration function
_migrate_add_testing_columns is currently a no-op but must add the deprecated
columns for DBs created after their removal; update _migrate_add_testing_columns
to check the target table schema (e.g., via inspector or PRAGMA) for the absence
of testing_in_progress and last_tested_at and, if missing, execute DDL to ALTER
TABLE and add testing_in_progress BOOLEAN NOT NULL DEFAULT FALSE and
last_tested_at DATETIME NULL (or equivalent for SQLite), ensuring the column
definitions match the SQLAlchemy model defaults and nullability and that the
operation is safe (wrap in try/except and commit/rollback as in existing
migration patterns). Ensure the function references the same table name used by
the model and handles both SQLite and other RDBMS dialect nuances when issuing
ALTER statements.

[coderabbitai]

Actionable comments posted: 1

🤖 Fix all issues with AI agents

In `@mcp_server/feature_mcp.py`:
- Around line 304-352: The feature_get_for_regression function currently reads
Feature.regression_count and then updates it in-memory, which allows lost
increments under concurrent requests; change the selection/update to perform
row-level locking or an atomic update: use
session.query(Feature).with_for_update().filter(Feature.passes ==
True).order_by(Feature.regression_count.asc(), Feature.id.asc()).limit(limit) to
lock rows before incrementing regression_count, then increment and commit (or
perform an UPDATE ... RETURNING to atomically increment regression_count and
fetch the rows), ensure session.commit() and session.refresh(...) are used after
the locked/atomic update so returned features reflect the new counts.

♻️ Duplicate comments (1)

api/database.py (1)

61-65: Missing migration for deprecated testing columns can break inserts.

Adding testing_in_progress / last_tested_at to the model means SQLAlchemy will include them in INSERTs; databases created after their removal but before this change won’t have those columns and will error. _migrate_add_testing_columns() is still a no‑op, so this gap persists.

🛠️ Suggested fix (in _migrate_add_testing_columns)

 def _migrate_add_testing_columns(engine) -> None:
-    """Legacy migration - no longer adds testing columns.
-
-    The testing_in_progress and last_tested_at columns were removed from the
-    Feature model as part of simplifying the testing agent architecture.
-    Multiple testing agents can now test the same feature concurrently
-    without coordination.
-
-    This function is kept for backwards compatibility but does nothing.
-    Existing databases with these columns will continue to work - the columns
-    are simply ignored.
-    """
-    pass
+    """Add testing columns if missing (backward compatibility)."""
+    with engine.connect() as conn:
+        result = conn.execute(text("PRAGMA table_info(features)"))
+        columns = [row[1] for row in result.fetchall()]
+        if "testing_in_progress" not in columns:
+            conn.execute(text(
+                "ALTER TABLE features ADD COLUMN testing_in_progress BOOLEAN DEFAULT 0"
+            ))
+        if "last_tested_at" not in columns:
+            conn.execute(text(
+                "ALTER TABLE features ADD COLUMN last_tested_at DATETIME DEFAULT NULL"
+            ))
+        conn.commit()

[pandel]

As you referenced my issue, I tested this PR:

• adding features via AI chat dialog worked now
• permissions to use WbeFetch or WebSearch are still not available