Framework method
Agent Skills
Focused instructions teach the agent Runable's APIs, conventions, and safe editing rules.
Runable × AIContext / 01
Framework-aware by design
AI should learn your framework before it edits your project.
Runable gives coding agents two things generic model knowledge cannot: precise framework instructions and the resolved state of the application in front of them.
Open formatRead onlyProject local
agent / runable
Request
Add a protected account page and load the current user during SSR.
load runable-pagesready
get_routes12 routes
get_extensionsauth found
app/pages/account.vuecreate
Uses existing auth middleware. Generated files stay untouched.
Method + factsContext complete
The missing context
01 / 04
A model may know Vue. It does not automatically know your Runable app.
Without explicit context, an agent can invent framework APIs, edit generated output, overlook existing middleware, or assume Runable owns the backend. More model intelligence does not fix missing project information.
01
No framework guessing
Use APIs and conventions that actually exist.
02
No blind file scanning
Read Runable's resolved view of the application.
03
No stale assumptions
Refresh the project state after a change.
Two sources of context
02 / 04
Instructions tell the agent how.
Inspection tells it what is true.
Project facts
Runable MCP
Read-only tools expose the routes, configuration, extensions, and diagnostics resolved right now.
“How should I build this?”
“How is this project built?”
Built-in coverage
03 / 04
Seven Skills
METHOD01
runable-projectUnderstand the project
02
runable-pagesBuild routes and layouts
03
runable-data-fetchingLoad SSR-aware data
04
runable-configurationConfigure and ship
05
runable-extendUse modules and plugins
06
runable-headManage SEO and metadata
07
runable-integrationsConnect the backend
Eight MCP tools
FACTSget_projectget_configget_routesget_extensionsresolve_routediagnosesearch_apirefreshstdio onlyread onlyprivate values withheld
Setup / 04
Add context to the project, not to every prompt.
Skills are committed with the codebase. The MCP server runs locally and inspects the Runable installation already used by the project.
01 / Skills
$ runable skills installSelect the coding agents used by your team, then commit the generated Skill directories.
02 / MCP
$ pnpm add -D @runablejs/mcp Connect your MCP client to the project-local runable mcp command.
Your agent, your choice
Use the coding tools already in your workflow.
Codex
Claude Code
Cursor
GitHub Copilot
Gemini CLI
Cline
Start with reliable context