Chasm

Universal Chat & Agentic AI System Manager
Harvest, merge, and recover your AI chat and agent task histories

Quick Start • Features • Installation • Docs • Contributing

Chasm bridges the divide between AI providers by extracting and unifying chat sessions from AI coding assistants like GitHub Copilot, Cursor, and more. Never lose your AI conversations again.

✨ Features

• 🔍 Harvest - Extract chat sessions from VS Code, Cursor, Windsurf, and other editors
• 🔄 Recover - Restore lost or orphaned chat sessions to VS Code
• 🔀 Merge - Combine sessions across workspaces and time periods
• 📊 Analyze - Get statistics on your AI assistant usage
• 🔌 API Server - REST API for building custom integrations
• 🤖 MCP Tools - Model Context Protocol support for AI agent integration
• 🗃️ Universal Database - SQLite-based storage that normalizes all providers

📦 Installation

From crates.io

cargo install chasm-cli

From source

git clone https://github.com/nervosys/chasm-cli.git
cd chasm-cli
cargo install --path .

Pre-built binaries

Download from GitHub Releases:

🚀 Quick Start

Recover Lost Chat Sessions

The most common use case - recover chat sessions that disappeared from VS Code:

# Recover sessions for a specific project
chasm fetch path /path/to/your/project

# Example output:
# [<] Fetching Chat History for: my-project
# ======================================================================
# Found 3 historical workspace(s)
#
#    [OK] Fetched: Implementing authentication system... (abc12345-...)
#    [OK] Fetched: Debugging API endpoints... (def67890-...)
#
# ======================================================================
# Fetched: 2 sessions
#
# [i] Reload VS Code (Ctrl+R) and check Chat history dropdown

After running, reload VS Code (Ctrl+R or Cmd+R) and your sessions will appear in the Chat history dropdown.

Auto-Detect Workspace Info

# See what chasm knows about a workspace
chasm detect all /path/to/your/project --verbose

# Output shows:
# - Workspace ID and status
# - Available sessions
# - Detected providers
# - Recommendations

List All Workspaces

chasm list workspaces

┌──────────────────┬──────────────────────────────────────────┬──────────┬───────────┐
│ Hash             │ Project Path                             │ Sessions │ Has Chats │
├──────────────────┼──────────────────────────────────────────┼──────────┼───────────┤
│ 91d41f3d61f1...  │ c:\dev\my-project                        │ 3        │ Yes       │
│ a2b3c4d5e6f7...  │ c:\dev\another-project                   │ 1        │ Yes       │
└──────────────────┴──────────────────────────────────────────┴──────────┴───────────┘

List Sessions for a Project

chasm list sessions --project-path /path/to/your/project

Search for Sessions

# Find sessions by project name
chasm find session "my-project"

# Find sessions containing specific text
chasm find session "authentication"

View Session Details

chasm show session <session-id>

Export Sessions

# Export to a backup directory
chasm export path /backup/dir /path/to/your/project

📖 Complete CLI Reference

Session Recovery & Fetching

Listing & Discovery

Viewing & Searching

Export & Import

Merging Sessions

Harvesting (Bulk Collection)

Git Integration

Provider Management

Server & API

🤖 Agency - Agentic Coding CLI

Chasm includes a full agentic coding toolkit similar to Claude Code CLI, but provider-agnostic. Run coding tasks with any LLM backend.

Available Tools

chasm agency tools

Agent Roles

chasm agency list

• coordinator - Orchestrates multi-agent workflows
• coder - Writes and refactors code
• reviewer - Reviews code for issues
• tester - Generates and runs tests
• researcher - Gathers information
• executor - Runs commands and tasks

Orchestration Modes

chasm agency modes

Usage Examples

# Simple coding task (single agent)
chasm agency run "Add error handling to main.rs"

# Specify a model
chasm agency run -m gpt-4o "Refactor this function to use async/await"

# Use local Ollama model
chasm agency run -m ollama/codellama "Write unit tests for lib.rs"

# Multi-agent swarm for complex tasks
chasm agency run --orchestration swarm "Build a REST API with authentication"

# Parallel agents for speed
chasm agency run --orchestration parallel "Analyze and fix all TODO comments"

🔄 Unified Chat Interface - No Vendor Lock-in

Chasm provides a unified interface to all chat systems, preventing vendor lock-in. Continue conversations seamlessly across providers.

The Problem

Your AI chat history is scattered across:

• VS Code Copilot (SQLite + JSON in workspaceStorage)
• Cursor (proprietary format)
• ChatGPT (web-only, export required)
• Claude (web-only)
• Local LLMs (various formats)

Each uses different formats, storage locations, and APIs. If you switch providers, you lose context.

The Solution

Chasm normalizes all sessions into a universal format and lets you:

1. Import from any provider into a unified database
2. Export to any format (JSON, Markdown, CSV)
3. Continue sessions with a different provider
4. Search across all history regardless of source

Continue a Session with Any Provider

# List all your sessions from all providers
chasm list sessions

# Export a Copilot session
chasm export sessions abc123 --format json --output session.json

# The exported session contains the full conversation:
# - All messages (user + assistant)
# - Tool invocations and results
# - Timestamps and metadata
# - File references and code blocks

# Continue the conversation with a different provider:
chasm agency run --context session.json "Continue implementing the feature"

# Or import into the harvest database for unified access
chasm harvest import session.json

Cross-Provider Workflow Example

# 1. Start a project with GitHub Copilot in VS Code
#    (sessions automatically tracked)

# 2. Later, recover and view your sessions
chasm fetch path /path/to/project
chasm list sessions --project-path /path/to/project

# 3. Export the session for portability
chasm export path ./backup /path/to/project

# 4. Continue with Claude, GPT-4, or local Ollama
chasm agency run -m claude-3 --context ./backup/session.json \
  "Review the code we wrote and suggest improvements"

# 5. Or merge multiple sessions into one unified history
chasm merge path /path/to/project

# 6. Search across ALL your AI conversations
chasm harvest search "authentication implementation"

Universal Session Format

Chasm's normalized format includes:

{
  "id": "uuid",
  "title": "Session title",
  "provider": "copilot|cursor|chatgpt|claude|ollama|...",
  "workspace": "/path/to/project",
  "created_at": "2026-01-08T12:00:00Z",
  "messages": [
    {
      "role": "user|assistant|system",
      "content": "Message text",
      "timestamp": "2026-01-08T12:00:00Z",
      "tool_calls": [...],
      "references": [...]
    }
  ],
  "metadata": {
    "model": "gpt-4o",
    "total_tokens": 15000,
    "files_referenced": ["src/main.rs", "Cargo.toml"]
  }
}

Benefits

🔌 API Server

Start the REST API server for integration with web/mobile apps:

chasm api serve --host 0.0.0.0 --port 8787

Endpoints

🗃️ Supported Providers

Editor-based

• ✅ GitHub Copilot (VS Code)
• ✅ Cursor
• ✅ Windsurf
• ✅ Continue.dev

Local LLMs

• ✅ Ollama
• ✅ LM Studio
• ✅ GPT4All
• ✅ LocalAI
• ✅ Jan.ai
• ✅ llama.cpp / llamafile
• ✅ vLLM
• ✅ Text Generation WebUI

Cloud APIs

• ✅ OpenAI / ChatGPT
• ✅ Anthropic / Claude
• ✅ Google / Gemini
• ✅ Azure AI Foundry
• ✅ Perplexity
• ✅ DeepSeek

📁 Database Locations

🛠️ Development

Prerequisites

• Rust 1.75+
• Git

Building

git clone https://github.com/nervosys/chasm-cli.git
cd chasm-cli
cargo build --release

Running tests

cargo test

📜 License

Licensed under the Apache License, Version 2.0 (LICENSE or http://www.apache.org/licenses/LICENSE-2.0).

🤝 Contributing

Contributions are welcome! Please read our Contributing Guide and Code of Conduct.

🔒 Security

For security issues, please see our Security Policy.

Security Audit Summary (v1.2.0)

Chasm underwent a comprehensive security audit in January 2026 against industry frameworks:

Key Security Features:

• 🔐 Argon2id password hashing (OWASP recommended)
• 🔑 JWT authentication with required secrets (no dev fallbacks)
• 🛡️ Parameterized SQL queries (no injection vectors)
• 🔒 DPAPI/Keychain integration for credential access

Dependencies: 2 transitive warnings from ratatui (TUI framework) - compile-time only, no runtime risk.

📞 Support

• 📖 Documentation
• 💬 GitHub Discussions
• 🐛 Issue Tracker
• 📧 Email Support

🤝 Contributing

We welcome contributions! Please see our Contributing Guide for details.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

📜 License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

_{Built with 🦀 Rust by Nervosys}

⭐ Star us on GitHub