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

# Deep Agents

Deep Agents checkpoints every thread through LangGraph's SQLite saver. A session is a thread in the root namespace; sub-agents run in their own checkpoint namespace inside the same database.

| | |
|---|---|
| **Runtime id** | `deepagents` |
| **Category** | LangGraph agent harness |
| **Vendor** | LangChain |
| **Tier** | Paid tier (Starter and up) |
| **Upstream** | [github.com/langchain-ai/deepagents](https://github.com/langchain-ai/deepagents) |

## Where the data comes from

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

- `~/.deepagents/.state/sessions.db` — the dcode CLI default
- `~/.deepagents/sessions.db` — legacy fallback

**Format.** LangGraph checkpoint schema: `checkpoints` and `writes`, both with serialised blobs.

## Environment variables

| Variable | Effect |
|---|---|
| `CLAWMETRY_DEEPAGENTS_DB` | point at any SqliteSaver database |

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

## Tokens and cost

**Tokens.** From each message's `usage_metadata`; a child's usage is its own, so per-child tokens never double-count.

**Cost.** Derived from the token split and model.

:::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

Namespaces written by the `task` tool are real delegations and become parented child sessions. Generic subgraph namespaces are surfaced only as metadata.

## Things worth knowing

- Note the `.state` subdirectory in the default path — the legacy path without it is only a fallback.
- There are two real message layouts (cumulative checkpoints and incremental writes) and they must be parsed as a union.
- Library users with no checkpointer persist nothing. Those runs never appear, and nothing is fabricated for them.

## Verify it is being read

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

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

# 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.