Warp native reference
Start with the Warp setup. Output keys are listed under output settings.
Native mappings
Rules
- Root rules go in
AGENTS.md. Scoped rules go to<scope>/AGENTS.md(tools that share it).
Skills
- Written to Warp skills at
.agents/skills/<name>/SKILL.md, the path Warp recommends. - A skill in a scope folder moves under that scope. Import keeps it there.
- Identical root copies are written once.
- Warp has no field that keeps a skill user-only. Agents use any skill when relevant, and
/{skill-name}runs it by hand. Sodisable-model-invocationis dropped with a note, and commands are skipped. - Warp also scans
.warp/skills/,.claude/skills/,.codex/skills/,.cursor/skills/,.gemini/skills/,.copilot/skills/,.factory/skills/,.github/skills/, and.opencode/skills/. agnostic-ai writes only.agents/skills/. WARP_SKILL_DIRSadds directories outside the repo for Cloud agents to index. It does not change which project folders Warp scans.
Workflows
- With
outputs.warp.workflows-dirset, each agent becomes a Warp Workflow at<dir>/<name>.yaml(name,command,description,tags). - Warp now recommends cloud Warp Drive workflows, but
{{path_to_git_repo}}/.warp/workflows/still loads. command:is the agent body as written, so write it as a shell snippet.x-warpsetsshells,arguments,source_url,author,author_url.import warpreads them back.
MCP
mcpServers has only the keys in Warp's MCP docs.
- Stdio:
command,args,env, andworking_directory(Warp'scwd).argsis required, so no arguments gives"args": []. - Remote (HTTP, SSE, WS):
urlandheaders, with notype, because Warp has no transport field. description,disabled, androotsare not written. Setdescriptionandrootsthroughx-warp.disabledgives a sync note, because Warp never starts project servers on its own.- Warp also loads Claude Code's
.mcp.json, Codex's.codex/config.toml, and.agents/.mcp.json, each after approval (provider table). - With
claudeorcodexsynced too, one server can sit in up to three files. Warp documents no handling of same-name servers, so approve one copy. import warprenamesworking_directoryback tocwd. An entry missingcommand(stdio) orurl(remote) is skipped, as in trae, antigravity, and windsurf.
Upgrading from WARP.md
- Warp recommends
AGENTS.mdbut still supportsWARP.md, which wins in a directory with both (Warp rules docs). - The first sync after upgrading renames a
WARP.mdthat agnostic-ai wrote at the configured root toWARP.md.bak. - A
WARP.mdyou wrote (noGenerated by agnostic-aimarker) is left alone. Sync warns that Warp misses the synced rules until you rename or remove it.
Global configuration
agnostic-ai sync --global --only warpadds home MCP specs to~/.warp/.mcp.json, with the same fields as project output.- The Warp app starts these servers automatically. The Agent CLI does not read this file (app and CLI).
- A spec with
disabled: trueis left out, with an explanation. agnostic-ai import --global warpreads the file and mapsworking_directoryback tocwd. See global MCP ownership and name conflicts.
Voice dictation
Add portable voice settings to your home specs, then run agnostic-ai sync --global --only warp. Sync writes enabled and language into Warp's settings file:
| System | File |
|---|---|
| macOS | ~/.warp/settings.toml |
| Linux | $XDG_CONFIG_HOME/warp-terminal/settings.toml, or ~/.config/warp-terminal/settings.toml when unset |
| Windows | %LOCALAPPDATA%\warp\Warp\config\settings.toml |
- Language is a lowercase two-letter code such as
es, or""for auto-detect. Broader values such asen-USandautoare skipped with a note. - Warp has no capture-mode setting, so
voice.modegives a note. The activation shortcut and other preferences stay in Warp. - Comments and unrelated keys in
[agents.voice]are kept. Removing portable voice restores the original values. - An existing inline or dotted definition of this table is left alone with an error asking for an explicit
[agents.voice]table. explain --global settings/voice.yamlshows the file and keys for your platform.
Warp app and Agent CLI
The app and the Agent CLI keep separate MCP settings (CLI configuration).
| Reader | Global MCP file | Project MCP file |
|---|---|---|
| Warp app | ~/.warp/.mcp.json | .warp/.mcp.json |
| Warp Agent CLI | its own file | .warp/.mcp.json, listed as off in /mcp until you enable each server |
- Global servers written by sync reach the app only.
sync --globaldoes not write the CLI file, so the CLI gets your rules and skills but no global MCP servers. - To copy your app setup to the CLI once, run
/tui-migrate-setupin the CLI. Later changes to~/.warp/.mcp.jsondo not follow. - CLI file path: Warp documents only macOS,
~/.warp_cli/.mcp.json. For other systems the path comes from Warp's source, not its docs:cli/.mcp.jsonin the Warp config directory. A running CLI has not confirmed it. The/mcpheader shows the file the CLI reads. - Warp's CLI page says project MCP files are not detected, but the shipped CLI lists servers from
.warp/.mcp.jsonin/mcp. They stay off until you enable each one.
Import
agnostic-ai import warp reads AGENTS.md, workflows from .warp/workflows/, and MCP servers from .warp/.mcp.json.
- Skills come in with all bundled files from
<root>/skills/for each root Warp scans:.agents,.warp,.claude,.codex,.cursor,.gemini,.copilot,.factory,.github,.opencode. - Within a scope, an earlier root wins a duplicate name. Scoped paths stay together.