---
title: Antigravity
description: How ClawMetry observes Antigravity — where its data lives, the environment variables that move it, what cost data is real, and the traps in its store.
eyebrow: Runtime reference
---

# Antigravity

Antigravity keeps one data root per product flavour under `~/.gemini`. Transcripts are plain JSONL, but token accounting lives in an undocumented protobuf blob in a per-conversation SQLite database — which ClawMetry decodes with a small wire walker rather than a protobuf dependency.

| | |
|---|---|
| **Runtime id** | `antigravity` |
| **Category** | Agentic IDE and CLI |
| **Vendor** | Google |
| **Tier** | Paid tier (Starter and up) |
| **Upstream** | [antigravity.google](https://antigravity.google) |

## Where the data comes from

ClawMetry reads these locations, in order, and never writes to any of them.

- `~/.gemini/{antigravity,antigravity-cli,antigravity-ide,jetski}/brain/<uuid>/.system_generated/logs/transcript_full.jsonl`
- `<root>/conversations/<uuid>.db` — `gen_metadata` token blobs (CLI flavour)
- `<root>/conversation_summaries.db` — parent/child conversation edges

**Format.** JSONL transcripts plus per-conversation SQLite carrying one protobuf blob per model generation.

## Environment variables

| Variable | Effect |
|---|---|
| `CLAWMETRY_ANTIGRAVITY_HOME` | data-root override |

See [Environment variables](/docs/config/environment/) for the variables that apply to every runtime.

## Tokens and cost

**Tokens.** Decoded from the `gen_metadata` protobuf: prompt, thinking and response counts, calibrated against live sessions.

**Cost.** Derived from the decoded token counts and the model id in the same blob.

:::note Why this section exists
Runtimes differ enormously in what they persist. Rather than showing a plausible number everywhere, ClawMetry reports what the store actually contains and marks the rest unknown. A cost of `$0.00` and a cost of *unknown* mean different things, and the dashboard keeps them different.
:::

## Sub-agents and delegation

Sub-agents are full conversations of their own. The only persisted delegation edge is the `parent_conversation_id` column in `conversation_summaries.db`.

## Things worth knowing

- `transcript_full.jsonl` is preferred over the truncated `transcript.jsonl`; reading the wrong one loses turns.
- Summary rows are WAL-buffered and can lag the brain directory, so a fresh child may briefly appear unlinked. ClawMetry never invents the missing edge.

## Verify it is being read

```bash
# Is this runtime detected, and would this install unlock it?
clawmetry runtimes --json | jq '.runtimes[] | select(.id=="antigravity")'

# Why is it locked, if it is?
clawmetry runtimes --why antigravity

# Any sessions ingested yet?
curl -s 'http://localhost:8900/api/local/sessions?limit=5' \
  | jq '.rows[] | {session_id, agent_id, events, cost_usd}'
```

If detection succeeds but no sessions appear, work through [Troubleshooting](/docs/ops/troubleshooting/) — the usual causes are a relocated data directory (set the override above), a session limit that is hiding older runs, or a runtime that simply has not written anything yet.