Skip to main content
Glama

Server Configuration

Describes the environment variables required to run the server.

NameRequiredDescriptionDefault
AGY_CMDYesAbsolute path to the agy binary

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": false
}
prompts
{
  "listChanged": false
}
resources
{
  "subscribe": false,
  "listChanged": false
}
experimental
{}

Tools

Functions exposed to the LLM to take actions

NameDescription
agy_run_startA

Start or continue one foreground Antigravity Run.

mode="task" starts a normal bridge-owned task. mode="interactive" starts a persistent conversation session that should be used sparingly. Supplying conversation_id continues that exact Antigravity conversation. Use agy_start_with_expected_file for reviews and other artifact-producing tasks. dangerously_skip_permissions must be true; the bridge always forwards --dangerously-skip-permissions to Antigravity.

agy_start_with_expected_fileC

Start one task Run that cannot complete until expected_file is non-empty.

agy_run_waitA

Wait for sparse Run events instead of repeatedly polling status.

run_ids is always a list, even for one Run. Supported condition values: any_attention, any_terminal, all_terminal, any_event, plus aliases attention, terminal, finished, finish, complete, completed, result, all_finished, all_complete, and all_completed.

agy_loginC

Refresh Antigravity auth state and optionally open one login session.

agy_run_observeA

Inspect Run state through one lean observation surface.

view="full" returns merged observable state for all run_ids. view="status", "transcript", or "terminal" requires exactly one run_id and returns the corresponding focused view.

agy_run_inputB

Send input to a live foreground Run with optional stale-write guards.

agy_run_cancelC

Cancel one active Run and terminate its Antigravity process group.

agy_run_resultB

Read final result metadata, or a bounded chunk when offset_bytes is set.

agy_review_commitA

Start a typed review Run for one commit and return immediately.

Keep issue focused and use narrow scope_paths when possible. After the run completes, prefer agy_review_result over manually polling raw artifacts.

agy_review_branchA

Start a typed review Run for branch and working-tree changes.

Keep issue focused and use narrow scope_paths when possible. Wait for completion with agy_run_wait, then prefer agy_review_result. Avoid frequent agy_run_observe(include_terminal_tail=True) calls unless debugging the bridge.

agy_review_filesC

Start an artifact-gated typed review Run for local files.

agy_review_resultA

Validate and summarize the artifact from a typed review Run.

Preferred way to consume completed commit, branch, or file review Runs.

agy_goalC

Manage bridge scheduler goals with actions create, start_target, status.

dangerously_skip_permissions must be true when supplied.

agy_adminC

Run bounded diagnostics and metadata actions for the bridge and agy CLI.

Prompts

Interactive templates invoked by user choice

NameDescription

No prompts

Resources

Contextual data attached and managed by the client

NameDescription

No resources

TDQS

B3.2/5.0

Scored across 14 tools

Disambiguation4/5

Most tools cleanly separate run lifecycle, review workflows, and bridge administration. The main ambiguity is that agy_run_start, agy_start_with_expected_file, agy_review_commit, agy_review_files, and agy_review_branch are all ways to start a run, and agy_run_result versus agy_review_result require careful reading.

Naming Consistency4/5

The shared agy_ prefix and snake_case convention make the set readable, and most tools follow a verb_noun style. Deviations like agy_start_with_expected_file, agy_admin, and agy_goal break the pattern slightly, but not enough to cause real confusion.

Tool Count5/5

Fourteen tools is a well-scoped surface for a bridge that manages run execution, review workflows, goals, login, and diagnostics. Each tool covers a distinct operation without feeling bloated or redundant.

Completeness4/5

The surface covers the full run lifecycle (start, wait, observe, input, cancel, result) and the main review workflows for commits, branches, and files. There are minor gaps such as no obvious way to enumerate all active runs without prior run IDs and limited goal mutation beyond status actions.

Maintenance

ActivityMaintained
ResponsivenessSlow