Skip to main content
Glama

Server Configuration

Describes the environment variables required to run the server.

NameRequiredDescriptionDefault
WEBCHATMCP_HOSTNoHTTP bind address (default `127.0.0.1`; `0.0.0.0` exposes to LAN — no auth, use with care)
WEBCHATMCP_PORTNoHTTP port (default `8321`; `0` disables HTTP)
WEBCHATMCP_CHANNELNoBrowser channel: `chromium` (default) / `chrome` / `msedge`
WEBCHATMCP_HEADLESSNo`1` runs headless (first login still needs a visible browser)
WEBCHATMCP_PROFILE_DIRNoBrowser profile directory (default `~/.webchatmcp/profile`)
WEBCHATMCP_ANSWER_TIMEOUT_MSNoAnswer wait limit (default `120000`)

Instructions

Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.

This server publishes no instructions, or was last inspected before Glama recorded them.

Capabilities

Features and capabilities supported by this server

Protocol revision2025-11-25

CapabilityDetails
tools
{
  "listChanged": true
}

Tools

Functions exposed to the LLM to take actions

NameDescription
webchat_loginA

Check whether the built-in browser profile is already logged in to the chosen service (headless, no window shown). If so, return immediately. Only when not logged in does it show a browser window so the user can log in manually once; the window is hidden again after login succeeds. The session is persisted in the profile directory and reused by later calls. Logging in is optional for chatgpt and gemini: webchat_ask also works as a guest there. Claude and Grok require it. Waits up to timeout_seconds (default 180) for the manual login; on timeout the window stays open and the user can finish later, then call webchat_status.

webchat_logoutA

Log out of the chosen service by clearing its cookies from the built-in browser profile (headless, no window shown), then re-check the login state. Logging out of gemini clears google.com cookies, which signs the profile out of Google. Use webchat_login to log in again. Guest use (where the service allows it) keeps working.

webchat_askA

Send a prompt through a fresh private/temporary chat of the chosen service in the built-in browser and return the answer text. Every call opens a brand-new chat: the prompt is not added to the account's chat history. Works without logging in (guest) on chatgpt and gemini; claude and grok require webchat_login (a guest gets a logged_out error). Optionally select a model first via model and then a thinking depth via thinking (labels as returned by webchat_models; pick thinking after model since the available depths can depend on the model; an unknown label fails with thinking_not_found). Prompts are typed exactly as given, including multi-line text.

webchat_modelsA

List the models and thinking-depth options available for the chosen service (scraped live from its model menu, so account tier differences are handled automatically). models entries have a label (use it with webchat_ask's model) and current. thinking lists the thinking depth / effort levels (ChatGPT slider, Claude effort, Gemini extended thinking) with current; it is empty for services without such a setting (Grok folds it into its modes). Works as a guest where the service shows a model menu.

webchat_statusB

Report whether the built-in browser is running, whether the login is active for the given service (defaults to the service of the current page), whether the current page is in private/temporary-chat mode, the profile directory, the HTTP endpoint state, and optional local CDP debugging endpoint (cdp.enabled / cdp.endpoint). Does not navigate: it inspects the page the browser is currently on.

webchat_closeA

Close the built-in browser cleanly. The login sessions stay saved in the profile directory and will be reused next time.

webchat_warmupA

Preload the private (temporary) chat page of the chosen service in the background browser, so the next webchat_ask skips the page load. With model (a label from webchat_models) the model is selected ahead of time too, and a later webchat_ask with the same model skips the switch. Meant for host integrations: call it when the user switches to a webchat model, and webchat_release when they switch away. Never an error when preloading is not possible (visible browser window, verification page, login required): warmed=false and the next webchat_ask just loads the page as usual.

webchat_releaseA

Close the background (headless) browser and its preloaded chat page because the user no longer uses any webchat model. The next webchat_ask or webchat_warmup starts it again. A visible browser window (login in progress) is left alone. Login sessions stay saved.

Prompts

Interactive templates invoked by user choice

NameDescription

No prompts

Resources

Contextual data attached and managed by the client

NameDescription

No resources

TDQS

A4.1/5.0

Scored across 8 tools

Disambiguation4/5

Most tools have clearly distinct purposes (login, logout, ask, models, status, warmup). However, webchat_release and webchat_close both shut down the browser and differ only subtly (release = stop using webchat models / keep preloaded page semantics vs close = full clean shutdown), which could cause misselection.

Naming Consistency5/5

All eight tools use a uniform webchat_verb pattern (login, logout, ask, models, status, warmup, release, close) with no stylistic deviations.

Tool Count5/5

Eight tools is well-scoped for a browser-driven webchat server, covering session, query, discovery, and lifecycle concerns without bloat.

Completeness4/5

The surface covers login lifecycle, querying, model discovery, and browser warmup/teardown, which fully supports the core workflow. Minor gaps exist, such as no explicit tool to pick or switch the target service or to continue a conversation, though the fresh-chat design makes these intentional.

Maintenance

ActivitySlowing
ResponsivenessNo recent community issues