Terminal-first IDE delivering RAG, indexing, and workspace intelligence to AI agents
A Rust-based CLI tool that gives AI agents deep workspace intelligence through the Model Context Protocol (MCP). Features intelligent code indexing, semantic search, LSP integration, session memory, and knowledge management.
Option 1: Install from Git (recommended)
cargo install --git https://github.com/ghatdev/agentideOption 2: Install from crates.io (coming soon)
cargo install agentideOption 3: Build from source
git clone https://github.com/ghatdev/agentide
cd agentide
cargo build --release
# Binary at: target/release/agentideBuild Requirements:
- Rust 1.70+
protobuf-compiler- Required at build time by LanceDB dependency- Ubuntu/Debian:
apt install protobuf-compiler - macOS:
brew install protobuf
- Ubuntu/Debian:
cd /path/to/your/project
agentide init
agentide indexCreates:
agentide.toml- Workspace configuration (at root).agentide/agentide.db- SQLite FTS5 database.agentide/embeddings/- LanceDB vector store
Workspace: init, status, file-tree, index
Search: search, search-regex, search-semantic, search-docs
Knowledge: Documentation and agent memory management
LSP: Language server lifecycle management (lsp start/stop/status/logs)
Symbols: symbols find, symbols references
Git: git-status, git-diff
Planning: workflow, plan, todos - Session-based task management with approval workflows
Configure in Claude Desktop (~/.config/Claude/claude_desktop_config.json):
{
"mcpServers": {
"agentide": {
"command": "/path/to/agentide",
"args": ["mcp"],
"cwd": "/path/to/workspace"
}
}
}Tool Categories:
- Workspace (2) - Status, file tree
- Search (5) - FTS5, regex, semantic, hybrid, docs
- Symbol Editing (5) - Find, replace, insert before/after
- Code Analysis (2) - File structure, diagnostics
- Knowledge (4) - Create, update, list, search
- Agent Lifecycle (5) - Sessions, insights, linking
- Planning (6) - Workflows, plans, todos
- RAG (1) - Embedding stats
- Smart Context (1) - Intelligent context gathering
See Claude Desktop Integration Guide
- SQLite FTS5 - Full-text search indexing
- Tree-sitter - 40+ language parsing
- FastEmbed - Local semantic embeddings (BGE-Small-EN-v1.5)
- LanceDB - Vector storage for similarity search
- Tower-LSP - Language server protocol integration
# Search code
agentide search "authenticate"
agentide search-semantic "user login flow"
# Task management
agentide workflow status
agentide plan list
agentide todos list --status pending
# Knowledge management
agentide knowledge create --title "API Design" --content "..."
agentide knowledge search "authentication"
# LSP operations
agentide lsp start rust
agentide symbols find AuthService
agentide diagnostics --file src/auth.rs
# Embeddings
agentide embed
agentide embedding-statsExample agentide.toml:
[workspace]
name = "My Project"
[repositories]
paths = ["."]
exclude_patterns = ["node_modules/", "target/", ".git/"]
[rag]
auto_index = true
auto_embedding = true
[[lsp.servers]]
language = "rust"
command = "rust-analyzer"
auto_start = true
[mcp_tools]
workspace = true
search = true
symbol_editing = true
knowledge = true
agent_lifecycle = true
smart_context = truecrates/
├── agentide/ # CLI binary & HTTP server (thin adapters)
├── agentide-core/ # Business operations (search, workflow, knowledge, etc.)
├── agentide-mcp/ # MCP server & 31 tools (thin wrappers)
├── agentide-parser/ # Tree-sitter parsing & code chunking
├── agentide-lsp/ # LSP client protocol
├── agentide-store/ # SQLite FTS5 + LanceDB (pure CRUD)
├── agentide-grammars/ # 44 native grammars + WASM support
└── agentide-types/ # Domain types + storage traits
- User Guide - Workflows and CLI reference
- Claude Desktop Setup - MCP configuration
- AGENTS.md - Quick reference for AI agents
# Build
cargo build --release
# Test
cargo test
# Run specific tests
cargo test --package agentide --test integration_tests
cargo test --package agentide --test mcp_unit_testsMIT OR Apache-2.0
- Issues: GitHub Issues
- Docs: docs/
Built for AI-assisted development