sessionbridge 0.1.3

MCP server for persistent cross-repo session memory with smart status tracking

SessionBridge

MCP server for persistent cross-repo session memory with smart status tracking

SessionBridge gives AI assistants like Bob persistent memory across conversations, with intelligent status tracking and event-driven progress saving.

Features

• 🧠 Persistent Memory - Sessions survive across conversations and context switches
• 📊 Smart Status Tracking - Automatically infers Done/Blocked/Next from your summaries
• 🎯 Task Goals - Every session starts with a clear objective
• 🔄 Multi-Repo Support - Track work across multiple codebases
• ⚡ Event-Driven - Save progress only when it matters
• 🎭 Optional - Works alongside normal conversations, not forced

Quick Start

Installation with uvx (Recommended)

# No installation needed! Run directly with uvx
uvx sessionbridge

Bob MCP Configuration

Add to your ~/.bob/mcp_settings.json:

{
  "mcpServers": {
    "sessionbridge": {
      "command": "uvx",
      "args": ["sessionbridge"],
      "alwaysAllow": [
        "list_sessions",
        "resume_session",
        "new_session",
        "save_progress",
        "end_session",
        "add_repos_to_session"
      ]
    }
  }
}

Restart Bob, and SessionBridge is ready!

Usage

Creating a Session

You: "Create a session called api-fix to fix the 500 errors in the user endpoint"
Bob: Creates session with goal and repo tracking

Working with Sessions

You: "I want to refactor the authentication system"
Bob: "Would you like me to:
      1. Create a new session to track this work?
      2. Continue an existing session?
      3. Just help without session tracking?"

Saving Progress

Bob automatically saves progress when:

• ✅ Root cause identified
• ✅ Architectural decision made
• ✅ Task completed
• ✅ Blocking issue hit
• ✅ Switching context

Resuming Work

You: "Resume the api-fix session"
Bob: Loads full context including status, progress, and next steps

Session File Format

Sessions are stored as markdown in ~/.sessionbridge/sessions/:

# api-fix
**Goal:** Fix 500 errors in user endpoint
**Repos:** /code/api-service, /code/database
**Created:** 2026-04-08

**Status:** in-progress
- Done: Identified root cause in query.py, Fixed N+1 problem
- Blocked on: Database migration approval
- Next: Deploy to staging, Run integration tests

## 2026-04-08
Identified root cause in database query. Fixed N+1 problem in user endpoint.
Blocked on database migration approval from DBA team.

**[Session ended at 18:30]**
Completed fix and local testing. Ready for staging deployment.

MCP Tools

1. new_session

Create a new session with task goal and repositories.

When to use: User explicitly wants session tracking for substantial work.

2. resume_session

Load session context. Optionally add new repositories.

When to use: Continuing previous work or expanding to new repos.

3. save_progress

Save significant progress with smart status inference.

When to use: Key findings, decisions, completions, blockers, or context switches.

4. end_session

End session with wrap-up summary.

When to use: Task complete, switching tasks, or handing off work.

5. list_sessions

List all sessions with last-modified dates.

When to use: Checking available sessions or helping user choose.

6. add_repos_to_session

Add repositories to existing session.

When to use: Work expands to additional codebases.

Smart Status Tracking

SessionBridge automatically extracts status from your summaries:

Summary: "Completed JWT validation. Blocked on API key - waiting for DevOps. 
          Next: integrate with API service."

Inferred Status:
- Status: blocked
- Done: JWT validation
- Blocked on: API key - waiting for DevOps
- Next: integrate with API service

Keywords

• Done: completed, finished, implemented, fixed, resolved, deployed
• Blocked: blocked on, waiting for, need, requires, pending
• Next: next, todo, need to, should, will, plan to

Advanced Features

Dynamic Repo Management

# Add repos when resuming
resume_session(name="api-fix", add_repos=["/code/frontend", "/code/mobile"])

# Or add explicitly
add_repos_to_session(name="api-fix", repos=["/code/admin"])

Auto-Trimming

Sessions automatically trim to last 20 date entries when token count exceeds 3000, keeping context manageable.

Installation (Development)

# Clone repository
git clone https://github.com/yourusername/sessionbridge
cd sessionbridge

# Install in development mode
pip install -e .

# Run directly
sessionbridge

Requirements

• Python 3.8+
• fastmcp

Why SessionBridge?

Problem: AI assistants lose context between conversations. You have to re-explain your project, what you've done, and what's next.

Solution: SessionBridge gives your AI assistant persistent memory with:

• Clear task goals
• Automatic status tracking
• Progress history
• Multi-repo awareness
• Event-driven saving (not noisy)

Result: Pick up exactly where you left off, even days later.

Best Practices

1. Use Sessions for Substantial Work

• ✅ Multi-day projects
• ✅ Cross-repo refactoring
• ✅ Complex bug investigations
• ❌ Quick questions
• ❌ One-off tasks

2. Write Meaningful Summaries

❌ Bad: "Made some changes"
✅ Good: "Implemented JWT refresh token logic in auth.py. Tested with 1000 concurrent users."

3. Let Sessions Be Optional

SessionBridge works alongside normal conversations. Use sessions when they add value, skip them when they don't.

Contributing

Contributions welcome! Please feel free to submit a Pull Request.

License

MIT License - see LICENSE file for details.

Authors

• Aakash Raghunathan ([e-mail])
• Tanmay Sreejith ([e-mail])

Links

• GitHub Repository
• Issue Tracker
• PyPI Package

---

Made with ❤️ for better AI assistant workflows