Zed Custom - Enhanced AI Agent Features

Important

UNOFFICIAL FORK: This project is a custom, unofficial fork of Zed and is not associated with, endorsed by, or affiliated with Zed Industries.

📖 Full Documentation

This is a custom fork of Zed with enhanced AI agent capabilities focused on improving code understanding and long-term memory.

🚀 Custom Features

1. LSP-Based Symbol Search

Status: ✅ Implemented

Replaced the original regex-based semantic indexing with native Language Server Protocol (LSP) integration for accurate, language-aware symbol navigation.

Benefits:

• ⚡ Faster: No CPU-intensive background indexing
• 🎯 More Accurate: Leverages language servers (rust-analyzer, pyright, gopls, etc.)
• 🔧 Type-Aware: Understands code structure, not just text patterns
• 🔄 Always Up-to-Date: Automatically maintained by LSP as you edit

Implementation:

• crates/agent/src/tools/context_tool.rs - New LSP-based context tool
• crates/agent/src/agent.rs - Disabled legacy regex indexing
• crates/agent/src/thread.rs - Updated tool initialization

Commit: 91b74d68c6

---

2. Agent Memory System: remember

Status: ✅ Implemented

The AI agent can proactively remember critical information using the built-in memory tools to store it in a persistent SQLite database tying knowledge directly to specific projects across editor restarts.

Example Usage:

User: "Remember that this project uses a microservices architecture with gRPC for inter-service communication"
Agent: Uses the 'remember' tool to permanently store this under the 'Architecture' category.

---

3. Agent Memory System: recall

Status: ✅ Implemented

The agent can retrieve previously stored memories using the recall tools to instantly pull up context stored days or weeks ago without needing to re-read files.

Example Usage:

[Later session]
User: "What architecture does this project use?"
Agent: Uses the 'recall' tool to retrieve the microservices architecture memory stored previously.

---

4. Instant Web Search & Deep Research

Status: ✅ Implemented

Integrated a full headless Chrome engine (chromiumoxide) for deep site analysis, while keeping a lightweight static scraper for quick searches.

Features:

• 🔍 Fast Search: Type /search <query> to instantly fetch DuckDuckGo results using a lightweight, no-JS scraper.
• 🧪 Deep Research: The /deep-research command uses headless Chrome to autonomously navigate, render, and extract Markdown from complex, JS-heavy websites (like React SPAs).
• 🔄 Smart Fallback: Both tools feature an automated fallback to static HTTP if the headless browser is unavailable, ensuring 100% reliability.

Example Usage:

User: "/search latest tailwind css features"
Agent: Fetches the search results and synthesizes the latest framework changes without leaving Zed.

---

5. Azure Anthropic & Token Caching UI

Status: ✅ Implemented

Enhanced support for Azure Anthropic deployments and natively enabled token caching visualization to help developers optimize their context reuse without manual config hacking.

Features:

• ☁️ Azure Ready: Passes exact serde_name string models to Azure APIs to resolve 404 DeploymentNotFound errors natively.
• 💾 Token Caching (Smart Strategy): Natively flags the system prompt and the second-to-last message in every turn. This ensures the entire conversation history remains "hot" in Anthropic's cache, resulting in <1s response times for massive projects.
• 📊 Caching UI: Enabled by default; automatically shows green cached and blue saved labels in the metadata row once the 1024-token threshold is crossed.

The "Smart Caching" Mechanism

To maximize the 5-minute cache window provided by Anthropic, we use a tiered strategy:

1. System Prompt: Always cached (Fixed).
2. Conversation Prefix: We dynamically flag the second-to-last message. This caches the bulk of your history while allowing the latest prompt to remain fresh.
3. Interval Pokes: Large threads are re-clamped every 15 messages to ensure long-term cache persistence.

References:

• Logic: crates/agent/src/thread.rs
• UI: crates/agent_ui/src/acp/thread_view.rs

Implementation:

• crates/anthropic/src/anthropic.rs - Added serde_name() context to Model enums
• crates/language_models/src/provider/anthropic.rs - Preserved model name transparently
• crates/agent_ui/src/agent_ui.rs - Flipped default config

---

🔧 Setup & Usage

Building from Source

# Clone this repository
git clone https://github.com/shotsan/zed-custom.git
cd zed-custom

# Build Zed
cargo build --release

# Run
cargo run

Syncing with Upstream Zed

This repository maintains zed-industries/zed as an upstream remote:

# Fetch latest from official Zed
git fetch upstream

# Merge upstream changes
git merge upstream/main

# Push to your fork
git push origin main

---

📝 Configuration Files

Custom project-specific rules can be defined in:

• .cpp_rules - C++ specific coding guidelines
• .python_rules - Python specific coding guidelines

---

🤝 Contributing

This is a personal fork with experimental features. If you find these features useful:

1. Consider contributing improvements via pull requests
2. Report issues specific to custom features
3. For core Zed issues, please report to zed-industries/zed

---

📜 License

Same as Zed - see their LICENSE files.

---

🔗 Links

• Upstream Zed: https://github.com/zed-industries/zed
• This Fork: https://github.com/shotsan/zed-custom

---

📊 Feature Comparison

Feature	Upstream Zed	This Fork
Symbol Search	Regex-based	✅ LSP-based
Agent Memory	❌ Session-only	✅ Persistent SQLite
Background Indexing	Heavy CPU usage	✅ Disabled (uses LSP)
Memory Categories	N/A	✅ 5 categories
Cross-Session Context	❌	✅ Yes
Web Search (/search)	❌	✅ Static DuckDuckGo
JS-Heavy Browsing	❌	✅ Headless Chrome
Azure Anthropic	❌ Needs proxy	✅ Transparently
Token Caching UI	❌ Hidden	✅ Enabled by Default

---

Last Updated: 2026-02-24

Zed

Welcome to Zed, a high-performance, multiplayer code editor from the creators of Atom and Tree-sitter.

---

Installation

On macOS, Linux, and Windows you can download Zed directly or install Zed via your local package manager (macOS/Linux/Windows).

Other platforms are not yet available:

• Web (tracking issue)

Developing Zed

• Building Zed for macOS
• Building Zed for Linux
• Building Zed for Windows

Contributing

See CONTRIBUTING.md for ways you can contribute to Zed.

Also... we're hiring! Check out our jobs page for open roles.

Licensing

License information for third party dependencies must be correctly provided for CI to pass.

We use cargo-about to automatically comply with open source licenses. If CI is failing, check the following:

• Is it showing a no license specified error for a crate you've created? If so, add publish = false under [package] in your crate's Cargo.toml.
• Is the error failed to satisfy license requirements for a dependency? If so, first determine what license the project has and whether this system is sufficient to comply with this license's requirements. If you're unsure, ask a lawyer. Once you've verified that this system is acceptable add the license's SPDX identifier to the accepted array in script/licenses/zed-licenses.toml.
• Is cargo-about unable to find the license for a dependency? If so, add a clarification field at the end of script/licenses/zed-licenses.toml, as specified in the cargo-about book.

---

📄 License & Affiliation

• Affiliation: This project is an independent fork and not affiliated with Zed Industries.
• Licensing:
  • All modifications in this fork are licensed under the MIT License.
  • The rest of the codebase follows the original Zed licensing: GPL-3.0 / AGPL-3.0 / Apache-2.0.