Zed combines conventional editor functions with several AI paths. For buyers, the important detail is where a request is executed and which component owns credentials: native Zed Agent, external agent over ACP, terminal-based agent, or edit prediction. Treating these paths as one product with one permission model creates false assumptions about security and cost.
Capabilities and integration boundaries
Zed’s documentation separates its native agent with file, terminal and MCP tools from external agents that use their own configuration and terminal threads that launch a CLI. Edit Prediction suggests text as the developer types; Inline Assistant transforms a selection. Agent Panel supports code changes and review checkpoints. Provider connectivity can use Zed-hosted models, API keys, gateways and local endpoints depending on the exact path.
A reproducible developer workflow
Test a small refactor in three separate sessions. First accept or reject an edit prediction while typing. Next ask Inline Assistant to rewrite a selected block and inspect the result. Finally ask native Zed Agent to refactor a single function while checking the tool-permission prompts and review UI. If testing an ACP agent, document its independent login and approvals rather than attributing them to Zed.
Limitations and failure cases
Edit prediction quality does not predict whether an autonomous agent can complete a multi-file change. A subscription that works for Zed-native models may not authenticate an external terminal agent. Nor does a local model option guarantee useful code completion under the team’s latency, memory and language requirements. Provider privacy statements have exceptions and should be read carefully.
Data handling, approvals and security
Disable unused model providers and MCP servers in sensitive projects. Confirm whether data is processed by Zed or by the underlying provider under the selected path, and check how feedback sharing works. Require human acceptance of edits and local regression testing. Use per-project settings rather than assuming a company-wide model policy covers external ACP tools.
Who should adopt it
Zed is particularly relevant to users who value a performant editor and want to switch between native, external and local AI workflows. Verify platform support and installed client version as part of any procurement decision.
Alternatives and decision trade-offs
Compare VS Code with configured AI extensions, JetBrains AI Assistant, and Cursor; compare the executing agent on its own merits rather than treating the editor as the entire product.
Version, region and entitlement caveats
Research and first-party documentation checked on 25 October 2026 (cutoff date). Documented version or status scope: not version-pinned. The documentation does not independently establish availability in every country, an exact installed client release, a universal enterprise contract, or a perpetual price. No unverified seat price, request quota, API rate, model entitlement or commercial-license guarantee is supplied here. Check AI 10 dated pricing snapshots and the signed agreement before procurement. The examples are suggested evaluation methods, not experiments or benchmarks performed for this article.
Checklist before production rollout
- Record the exact installed version, service account, organization policy and contract or plan applicable to the tested environment.
- Create a clean worktree; specify the task boundary, forbidden paths, acceptance criteria and rollback before tool execution.
- Audit file, shell, network, extension and repository-app permissions; deny all unnecessary credentials and privileged actions.
- Review each resulting diff for unrequested packages, automation changes, data leaks and behavioral regressions.
- Run targeted tests, existing integration/regression tests and independent code/security review in a trusted environment.
Sources & Further Reading
- https://zed.dev/docs/ai/overview (official publisher documentation; checked 25 October 2026)
- https://zed.dev/docs/ai/zed-agent (official publisher documentation; checked 25 October 2026)
- https://zed.dev/docs/ai/edit-prediction (official publisher documentation; checked 25 October 2026)
- https://zed.dev/docs/ai/llm-providers (official publisher documentation; checked 25 October 2026)
- https://zed.dev/privacy (official publisher documentation; checked 25 October 2026)