What are Zed Terminal Threads?

Zed Terminal Threads are a feature in Zed 1.3.5 (released May 20, 2026) that let you run any terminal-based AI coding agent — Claude Code, OpenAI Codex CLI, Amp, or any custom CLI agent — as a thread inside Zed's agent panel sidebar. Each thread can use a different agent, and you can run Zed's own built-in chat agent in one thread while running external terminal agents in others. All threads share editor context (current file, selection, project root).

Why does Zed support multiple agents instead of just one?

In May 2026, no single AI coding agent is best at everything. Claude Opus 4.7 wins SWE-bench. GPT-5.5 wins certain reasoning tasks. Gemini 3.5 Flash wins on cost. Cursor's Composer 2.5 is great for cheap routine work. By supporting any terminal agent as a thread, Zed lets developers mix and match per task without leaving the editor. Zed positions itself as the meta-routing IDE — the editor for power users who refuse single-vendor lock-in.

How is Zed different from Cursor?

Cursor is an opinionated AI IDE — it uses Cursor's model router with Composer 2.5 as default and Claude/GPT/Gemini routed through Cursor's billing. Zed is editor-agnostic about agents — you bring your own terminal agent (Claude Code, Codex, Amp, etc.) and pay your vendor directly. Cursor wins on out-of-box polish and single-bill simplicity. Zed wins on flexibility, speed (Rust-native, faster than VS Code/Cursor), being free, and avoiding lock-in.

Yes, Zed itself is free and open-source. You only pay for the AI agents you run inside it — Claude Code (Anthropic billing), Codex (OpenAI billing), or direct API usage through whatever provider you choose. Combined with direct API billing for agents (and using prompt caching), Zed + Terminal Threads is often the cheapest serious AI coding setup in May 2026, especially after Anthropic's June 15 Claude Code billing change makes direct API more attractive.

Quick Answer

What is Zed Terminal Threads? Meta-IDE for AI Agents (May 2026)

Published: May 28, 2026

What is Zed Terminal Threads? Meta-IDE for AI Agents (May 2026)

Zed Terminal Threads shipped on May 20, 2026 in Zed 1.3.5. They turn Zed into the meta-routing IDE — a fast, free editor where you can run multiple AI coding agents from different vendors as threads in one panel. Here’s what it is, why it matters, and how to use it.

Last verified: May 28, 2026.

TL;DR

What: A feature in Zed’s agent panel that lets you run terminal-based AI agents (Claude Code, Codex CLI, Amp, custom CLIs) as threads alongside Zed’s built-in chat agent.
Released: May 20, 2026 in Zed 1.3.5
Tagline: “Run Claude Code, Amp, or any terminal agent as threads in Zed’s sidebar.”
Cost: Zed is free; you pay agent vendors directly.
Best for: Power users who refuse single-vendor lock-in.

What Zed is in May 2026

Zed is a fast, open-source code editor built in Rust by the same team that built Atom. It’s been gaining traction as a serious alternative to VS Code and Cursor since 2023, prized for:

Speed — Rust-native, GPU-accelerated rendering, multi-process architecture
Collaboration — built-in multi-cursor multiplayer editing
AI integration — agent panel with chat-style conversations
Open source — GPL-licensed, community-developed

The agent panel originally hosted only Zed’s own chat-style AI assistant. Terminal Threads extends it.

How Terminal Threads work

From Zed’s docs (zed.dev/docs/ai/agent-panel):

“The Agent Panel can host terminal threads, letting you switch between conversations and shell sessions from the same list. External agents like Claude Agent and Codex can also run as terminal threads.”

And from zed.dev/docs/ai/parallel-agents:

“Each thread can use a different agent, so you can run Zed’s built-in agent in one thread and an external agent like Claude Code or Codex in another.”

The shape:

Open Zed’s agent panel (right sidebar)
Create a new thread — choose chat or terminal
For terminal threads, choose an agent:
- Zed’s built-in chat agent
- Claude Code (Anthropic) — requires claude-code CLI installed
- OpenAI Codex CLI — requires codex CLI installed
- Amp — requires Amp CLI installed
- Any custom CLI that follows the agent protocol
Threads share editor context — current file, selection, project root are accessible
Switch between threads without losing state

Why this matters in May 2026

The AI coding agent market has fragmented. As of May 2026:

Agent	Strength
Claude Code	Best SWE-bench, terminal-native, subagent orchestration
OpenAI Codex	Best mobile remote control (ChatGPT app), cloud parallelism
Cursor Composer 2.5	Cheapest fast default for routine work
Devin (Cognition)	Best autonomous remote agent for backlogs
Gemini 3.5 Flash	Cheapest multimodal fast frontier model
Amp	Smaller, but interesting for some terminal flows

No single agent is best at everything. Smart developers want different agents per task. Until Zed Terminal Threads, that meant juggling multiple terminal windows, multiple editors, or copy-pasting between systems.

Zed Terminal Threads collapse this into one panel.

Concrete workflow examples

Example 1: SWE-bench-class hard bug + simple test fix

Zed agent panel:
├── Thread 1: Claude Code (Opus 4.7)
│     "Investigate this race condition in connection_pool.py..."
├── Thread 2: Codex CLI (GPT-5.5)
│     "Add tests for the new auth middleware..."
└── Thread 3: Zed chat (Gemini 3.5 Flash)
      "Quick question: what does this regex match?"

Three different agents, three different cost profiles, one editor.

Example 2: Cost-optimized routine work

Zed agent panel:
├── Thread 1: Codex CLI (GPT-5.5 mini)
│     Routine refactor across 12 files (cheap)
└── Thread 2: Claude Code (Sonnet 4.6)
      Code review and improvement suggestions (cheap)

Pay only for cheaper models when the task doesn’t need a frontier model.

Example 3: Multi-vendor comparison

Zed agent panel:
├── Thread 1: Claude Code (Opus 4.7) — same task
└── Thread 2: Codex CLI (GPT-5.5) — same task

(Compare outputs, pick the better implementation)

This is genuinely valuable for hard problems where you want a second opinion from a different model family.

Pricing reality check

Zed itself is free. You pay for whatever agents you run:

Agent	Pricing
Claude Code	Claude Pro $20 / Max 5x $100 / Max 20x $200 (billing splits June 15, 2026)
OpenAI Codex	ChatGPT Plus $20+ for desktop / mobile control; API for direct
Amp	Direct API
Custom CLI	Whatever you wire up

Pro tip: after Anthropic’s June 15, 2026 Claude Code billing change (separate credit pool at full API rates, no rollover), heavy users will want to switch to direct Anthropic API billing. Zed Terminal Threads work with any auth method — set ANTHROPIC_API_KEY and Claude Code in Zed uses your direct API account, not your Pro/Max subscription pool.

How this compares to Cursor 3 Agents Window

Cursor 3 Agents Window (April 2026) also supports parallel agents. The difference:

	Zed Terminal Threads	Cursor 3 Agents Window
Editor	Zed (Rust, fast, free)	Cursor (VS Code fork, $20-$200/mo)
Agent vendor	Bring your own (any terminal agent)	Cursor’s router (Composer 2.5 default, others on tap)
Billing	Direct to agent vendor	Bundled in Cursor subscription + overage
Lock-in	None	Cursor ecosystem
Polish	More configuration needed	Out-of-box polished

Zed wins on flexibility and cost. Cursor wins on single-bill simplicity and out-of-box polish.

Who should use Zed Terminal Threads

Great fit

Senior engineers who want zero vendor lock-in
Cost-optimizers running heavy AI usage who want direct API billing
Power users comfortable installing multiple CLI agents
Anti-Cursor folks (philosophical or pricing-driven)
Linux users who never adopted Cursor
Multi-vendor enterprises that want flexibility

Not a great fit

Beginners — Cursor Pro $20 is more out-of-box friendly
Single-vendor shops committed to Cursor or VS Code + Copilot
Mobile-first workflows — Zed has no mobile counterpart
Teams that prioritize unified bill — direct API to multiple vendors means multiple bills

What’s next for Zed

Reasonable predictions:

More agents shipping native integration — every major terminal AI agent will want to be a first-class Zed thread
Better thread orchestration — running threads in sequence, passing context between threads
Possible cloud agent support — Cursor 3 has cloud agents; Zed will likely follow
Larger market share among power users — Terminal Threads is the kind of feature that converts senior engineers permanently

Verdict

Zed Terminal Threads (May 20, 2026, Zed 1.3.5) is the most important editor feature of Q2 2026 for power users. It re-frames Zed from “fast alternative to VS Code” to “the meta-IDE for AI agent workflows.”

If you’re tired of Cursor’s opinionated routing, or want to run Claude Code and OpenAI Codex side-by-side, or just want to pay agent vendors directly without an editor-tier markup, Zed + Terminal Threads is now the obvious choice.

For beginners and single-vendor shops, Cursor Pro is still the easier path. For everyone else, give Zed 1.3.5 a serious try.

Sources: Zed Release Notes May 2026, Zed Agent Panel documentation (zed.dev/docs/ai/agent-panel), Zed Parallel Agents documentation (zed.dev/docs/ai/parallel-agents), Anthropic billing change announcement (May 13, 2026), byteiota.com Zed Terminal Threads guide, digitalapplied.com AI Agent Stack Decision Tree.