Cortex Browser MCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| CORTEX_ALLOWED_ORIGINS | Yes | Comma-separated list of exact allowed origins, including scheme, host, and port, without trailing slash. No wildcards. Browser calls are refused without valid origins. |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
Usa cortex_open con newSession=true, conserva sessionId, peek mirati e cortex_run in batch. Se presenti nel catalogo: observe per pixel fedeli, inspect per layout, capture prima delle azioni. JavaScript/DevTools solo con capability host. Chiudi la tua sessione con cortex_debug(closeSession=true). Dopo timeout consulta operationId/stato, nessun replay automatico.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| cortex_openC | Apri URL autorizzato nella sessione condivisa; nessuno snapshot automatico. |
| cortex_peekB | Controlli/griglie o estrazione mirata. Usa i ref restituiti; dopo navigazione o sostituzione AJAX aggiorna quelli necessari. scope opzionale: CSS di un solo contenitore verificato. items: CSS dei record ripetuti. fields maxLength 1..500 (default 300). Niente HTML o valori input. |
| cortex_runB | Azioni/assert e piani each/expand/extract. Target: ref:s1-e1 o s1-e1; CSS; role:button|Nome, label:Nome. each/extract solo CSS; each limit max 50; fields maxLength 1..500. 25 step/200 azioni/60s. StaleRef/AmbiguousTarget: step non eseguito, max 5 candidati correnti; scegli il controllo e riprendi dallo step fallito senza ripetere quelli completati. |
| cortex_debugC | Diagnostica compatta senza contenuti console/rete; screenshot locale solo con screenshot=true. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 4 tools
The four tools map to distinct phases: open navigates, peek reads/extracts, run performs actions/asserts, and debug diagnoses. However, peek and run both support extraction, so an agent could reasonably hesitate between them when the task is only to extract data.
All tools use the same cortex_ prefix followed by a short lowercase verb. The pattern is fully predictable across open, peek, run, and debug.
Four tools is lean but appropriate for a focused browser automation server. Each tool covers a distinct interaction phase and none appears redundant.
The set covers the core browser loop: navigation, inspection/extraction, actions/assertions, and diagnostics. Minor gaps exist around explicit session close/history or standalone screenshot control, but they are workable through the provided tools.