Pass an explicit model to every spawn_agent call and add BYOK keys for Opus 5.5 and GPT-6
Zed v1.22.0-pre adds a per-call model on spawn_agent, automatic subagent compaction, and BYOK for Claude Opus 5.5 plus GPT-6 Astra, Sol, and Luna. Stop inheriting the parent model; wire your own Anthropic/OpenAI keys and route each spawn on purpose.Why this loop
The workflow change in this prerelease is model routing, not a new agent UI. spawn_agent now accepts an optional model so a spawn can differ from the configured subagent model or the parent, and the active model is shown on its card. BYOK unlocks Claude Opus 5.5 (Anthropic key) and GPT-6 Astra, Sol, and Luna (OpenAI key), so a frontier parent can spawn cheaper or specialized workers. Subagent compaction is enabled, so long-running workers no longer need hand-written summarization. Also in this build: Grok 4.7 is the SuperGrok/xAI recommended default, OpenCode Zen/Go models are fetched dynamically, and agent.threads_sidebar.auto_open controls whether opening a folder in an existing window opens the Threads Sidebar. Do not infer the 20 omitted changelog items.
Proposed actions
- Upgrade Zed to v1.22.0-pre from https://github.com/zed-industries/zed/releases/tag/v1.22.0-pre, then in Agent settings add an Anthropic API key and enable BYOK for Claude Opus 5.5 on parent coding threads.
- In Zed Agent settings, add an OpenAI API key and enable BYOK for GPT-6 Astra, GPT-6 Sol, and GPT-6 Luna; assign those models to spawned workers instead of the parent.
- Change every spawn_agent invocation to pass an explicit model argument instead of using the configured subagent model or inheriting the parent; confirm the chosen model on the subagent card.
- Remove manual compaction or summarization prompts from subagent instructions — Zed now compacts subagents automatically — and let long-running spawns run until the task finishes.
- If you use SuperGrok or xAI, select Grok 4.7 as the provider default (it is now the recommended default). Set agent.threads_sidebar.auto_open to false in settings.json if opening a folder in an existing window should not also open the Threads Sidebar.
Agent prompt
Paste into your agent or query via MCP (get_agent_prompt) — free, no extra AI costAGENTS.md / CLAUDE.md / GEMINI.md rule update
DevAgentRadar → Repo Harness Rule Patch
Goal: update our repository's permanent harness instructions based on this release.
Update the file that tool actually reads. Do not dump everything into one file.
- Claude Code → CLAUDE.md plus
.claude/(skills, hooks, settings, agents, commands). It does not read AGENTS.md natively; start CLAUDE.md with@AGENTS.md. - Codex, Copilot, Cursor, Factory Droid, Grok Build, Roo Code, Goose, OpenCode, Amp, Zed, Aider → AGENTS.md.
- Gemini CLI / Antigravity → GEMINI.md. AGENTS.md only if
context.fileNameis set. - Cursor glob-scoped rules →
.cursor/rules/*.mdc(plain.mdis ignored), not a second constitution. - Codex MCP →
.codex/config.toml, not.mcp.json. - Copilot extra instructions →
.github/copilot-instructions.md; custom agents →.github/agents/. - Windsurf →
.windsurf/rules(do not assume AGENTS.md). - Cline →
.clinerules. - Procedures → a skill (
SKILL.md). Enforcement the model must not skip → a hook. Isolated roles → subagents. - Do not fork the same rule into three tool files.
Context
Assistant: Zed AI Proposal: Pass an explicit model to every spawn_agent call and add BYOK keys for Opus 5.5 and GPT-6 Summary: Zed v1.22.0-pre adds a per-call model on spawn_agent, automatic subagent compaction, and BYOK for Claude Opus 5.5 plus GPT-6 Astra, Sol, and Luna. Stop inheriting the parent model; wire your own Anthropic/OpenAI keys and route each spawn on purpose. Primary source: https://github.com/zed-industries/zed/releases/tag/v1.22.0-pre
Why it matters
The workflow change in this prerelease is model routing, not a new agent UI. spawn_agent now accepts an optional model so a spawn can differ from the configured subagent model or the parent, and the active model is shown on its card. BYOK unlocks Claude Opus 5.5 (Anthropic key) and GPT-6 Astra, Sol, and Luna (OpenAI key), so a frontier parent can spawn cheaper or specialized workers. Subagent compaction is enabled, so long-running workers no longer need hand-written summarization. Also in this build: Grok 4.7 is the SuperGrok/xAI recommended default, OpenCode Zen/Go models are fetched dynamically, and agent.threads_sidebar.auto_open controls whether opening a folder in an existing window opens the Threads Sidebar. Do not infer the 20 omitted changelog items.
Suggested actions
- Upgrade Zed to v1.22.0-pre from https://github.com/zed-industries/zed/releases/tag/v1.22.0-pre, then in Agent settings add an Anthropic API key and enable BYOK for Claude Opus 5.5 on parent coding threads.
- In Zed Agent settings, add an OpenAI API key and enable BYOK for GPT-6 Astra, GPT-6 Sol, and GPT-6 Luna; assign those models to spawned workers instead of the parent.
- Change every spawn_agent invocation to pass an explicit model argument instead of using the configured subagent model or inheriting the parent; confirm the chosen model on the subagent card.
- Remove manual compaction or summarization prompts from subagent instructions — Zed now compacts subagents automatically — and let long-running spawns run until the task finishes.
- If you use SuperGrok or xAI, select Grok 4.7 as the provider default (it is now the recommended default). Set agent.threads_sidebar.auto_open to false in settings.json if opening a folder in an existing window should not also open the Threads Sidebar.
Config surfaces this release may change
- context and compaction settings — check your repo before applying
- subagent definitions — check your repo before applying
After you finish
Do not report this as applied to DevAgentRadar. You cannot write the visitor's loop.
Tell the human: open https://devagentradar.com/proposals/zed-ai-v1-22-0-pre-pass-an-explicit-model-to-every-spawn-agent-call-and- and mark Applied, Skipped, or Failed. Proposal id: 1155219e-6e9d-4207-bbd8-24141953b92c
Instructions:
- Read the existing context files and settings for the tools in this repo.
- Draft an explicit Git diff. Update ONLY agent harness rules or tool configuration.
- Call out deprecated flags, obsolete habits, or changed permission boundaries.
- Do not touch application logic.
Start with the proposed diff and say which layer and which tool's file it belongs in.
Your loop
This browser · no sign-in · not shared as “you”After you run the prompt
Only you can mark this. Agents cannot write your loop.
Your decision stays on this device. A public tally appears after a few votes.