Respira WordPress Playbooks
The agent authors a workflow once. Future runs are one tool call.
A Playbook is a JSON composition that registers itself as a WordPress Ability and shows up in the MCP tool catalog. The agent describes the workflow, Respira stores it, and future invocations are a single typed call instead of an instruction-by-instruction agent run. Production-safe by construction: destructive operations are refused at create-time, every step runs through Respira’s existing audit and snapshot belt.
The Respira playbook atlas · opt-in directory · empty until customers share theirs
Three nouns. One agent. Zero PHP.
Respira’s v7.1 surface stacks cleanly. Pick the right tool by what the agent is doing, not by which file lives where.
The agent reads.
Markdown documentation the agent loads into context. 35 skills today
(case study workflows, migrations, audits). Distributed via the agent-
skills-wordpress repo; installed via npx @respira/wordpress-mcp-server install-skills.
The agent runs.
JSON workflow compositions the agent authors at runtime. Each registers as a WordPress Ability and becomes a callable MCP tool. One tool call, one typed result, ~200 bytes of agent context per invocation.
See the directory ↓The WordPress primitives.
The wp_register_ability core API. Any plugin can expose abilities; Respira absorbs them through Inhale. The universal plug under everything — Playbooks themselves are abilities.
Browse abilities →Production-safe by construction.
Other AI integrations
Give the agent execute_php + filesystem write. The agent
generates a PHP file, drops it in wp-content/plugins/,
WordPress loads it on the next request. Powerful. No sandbox, no
approval gate, no audit trail. Vendor footers admit it: “for development
and staging environments only.”
Respira Playbooks
The agent crystallizes the workflow into a typed JSON composition. Respira stores it, registers it as a WordPress Ability, executes each step through the existing duplicate-before-edit + audit-log + snapshot belt. Destructive tools are refused at create-time — split the flow. Production-safe. Auditable. Reversible. Customer can delete any Playbook anytime from wp-admin.
Public directory
Playbooks customers opted in to publish. Browse them, read the source (Playbooks are pure JSON), copy them to your own site via export/import. Last refreshed: 5/30/2026.
The directory crawl hasn’t populated yet on this deploy. The v7.1 release is the first one to ship Playbooks; the public listing fills in as customers opt to publish.
In the meantime, you can browse the agent-readable directory JSON:
curl https://www.respira.press/playbooks/data.json
Teach your agent
Have a Respira-connected agent (Claude Code, Cursor, Codex, Windsurf, any MCP client). Paste this into the chat. The agent will use the Build-a- Playbook skill if installed, or fetch its source from the agent-skills repo if not.
Use the build-a-playbook Respira skill on this site. I want to crystallize this workflow into a reusable tool so future runs are one call: [describe the workflow in plain English] After you've designed the input schema and steps, confirm the playbook id and label with me before calling respira_create_playbook. I want to review the JSON before it lands.
After the agent confirms, the new Playbook lands in your site’s
wp-admin under Respira → Playbooks and shows up in the MCP tool catalog
on the next handshake as respira-playbooks/<id>.
