Skip to main content
Glama

Server Configuration

Describes the environment variables required to run the server.

NameRequiredDescriptionDefault
TOUCH_BROWSER_SERVE_COMMANDNoForce a specific binary or wrapper command for the serve operation
TOUCH_BROWSER_EVIDENCE_NLI_MODEL_PATHNoOverride the default location for the evidence NLI model
TOUCH_BROWSER_EVIDENCE_EMBEDDING_MODEL_PATHNoOverride the default location for the evidence embedding model

Capabilities

Features and capabilities supported by this server

CapabilityDetails
tools
{
  "listChanged": false
}

Tools

Functions exposed to the LLM to take actions

NameDescription
tb_statusA

Return runtime and daemon capability status.

tb_cancelA

Best-effort cancellation and daemon reset for long-running MCP browser work. MCP hosts should prefer notifications/cancelled for an in-flight request.

tb_session_createA

Create a headless research session for public docs and reference workflows. The caller may provide sessionId as a correlation id; otherwise the daemon generates one. The returned active tab is empty until tb_open or tb_search_open_top opens a page.

tb_openB

Open a public web document or official reference page headlessly, either statelessly or inside a daemon session/tab. target is required for stateless use; with sessionId it can be omitted to reopen the active tab URL.

tb_searchA

Search the public docs/research web headlessly and structure the result page for follow-up browsing. Search engine selection is automatic; follow tb_search_open_top before read-view or extract.

tb_search_open_resultA

Open one structured search result into a new research tab when you must override the recommended order after tb_search.

tb_search_open_topA

Open the top recommended search results into new research tabs. Prefer this immediately after tb_search, then inspect tabs with tb_read_view before tb_extract.

tb_extractA

Extract claim outcomes from a target or opened daemon tab. In a session, omit target to use the active opened tab; claims is always required.

tb_read_viewA

Return a readable Markdown view of a target or opened daemon tab for scope checking. In a session, omit target to use the active opened tab.

tb_policyB

Return the policy evaluation for a target or daemon tab on the public docs/research web surface.

tb_tab_openB

Create a new headless research tab, optionally opening a narrowed public web target immediately.

tb_tab_listB

List all daemon tabs registered for a session.

tb_tab_selectC

Set the active daemon tab for a session.

tb_tab_closeC

Close one daemon tab and update the active tab if needed.

tb_checkpointB

Return the current supervised checkpoint guidance for a daemon session/tab.

tb_profileB

Return the active policy profile for a daemon session/tab.

tb_profile_setC

Set the active policy profile for a daemon session/tab.

tb_clickB

Click an interactive target inside an existing daemon session/tab.

tb_typeC

Type into an interactive field inside an existing daemon session/tab.

tb_approveB

Persist supervised approval risks for the current daemon session so repeated ack flags are not required.

tb_secret_storeC

Store a sensitive value only in daemon memory for a specific target ref.

tb_secret_clearA

Clear one stored daemon secret or all secrets for a session.

tb_type_secretA

Type a previously stored daemon secret into a sensitive field without persisting it to disk.

tb_submitC

Submit a form or submit control inside an existing daemon session/tab.

tb_refreshC

Refresh the current headless research tab after query or filter changes.

tb_telemetry_summaryB

Return the aggregated pilot telemetry summary for the current runtime.

tb_telemetry_recentC

Return recent pilot telemetry events for the current runtime.

tb_session_synthesizeA

Aggregate opened tabs inside a daemon session into a citation-ready synthesis report. Requires sessionId and at least one opened tab.

tb_session_closeA

Close a daemon session and clean up all tab state.

Prompts

Interactive templates invoked by user choice

NameDescription

No prompts

Resources

Contextual data attached and managed by the client

NameDescription

No resources

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/nangman-infra/touch-browser'

If you have feedback or need assistance with the MCP directory API, please join our Discord server