Zed capabilities
10 mapped capabilities, each graded and dated. The map shows what Zed can do; the audit shows whether it’s worth consolidating — and a guide shows how to move.
Capabilities
Agent Profiles and fine-grained tool permissions
provisionalverified yesterdayGranular control over which tools an agent thread can use, configured through reusable profiles.
Agentic editing in the Agent Panel
provisionalverified yesterdayAn AI agent reads, writes, and runs code in your project: you describe a task and the agent discovers context, edits files, and runs commands until the task is done.
Bring your own keys / external LLM providers
provisionalverified yesterdayUse your own LLM provider accounts, subscriptions, gateways, or local models instead of Zed-hosted billing.
Edit Prediction (next-edit autocomplete, Zeta)
provisionalverified yesterdayAn LLM predicts the code you want to write next at keystroke granularity, suggesting individual or multi-line edits you accept by pressing tab.
Editable unified-diff review of agent changes
provisionalverified yesterdayBefore committing, you review the agent's work as an editable unified diff and accept or reject the changes.
External agents via Agent Client Protocol (ACP)
provisionalverified yesterdayRun third-party coding agents inside Zed's Agent Panel through a standard protocol, with the external agent owning its own runtime, auth, model, and tools.
MCP context servers
provisionalverified yesterdayExtend the agent's knowledge and tools by connecting external tools and data sources through the Model Context Protocol.
Native GPU-accelerated code editor
provisionalverified yesterdayA high-performance, open-source code editor built from scratch in Rust, designed for low-latency editing and responsive AI interactions.
Plans and pricing
provisionalverified yesterdayZed's subscription tiers for its AI features and hosted-model billing.