embed DGC in an application or CI job

The DGC SDK embeds the same agent in your own process. It is free and local. You do not pay DGC to use it. Optional Pricing on a client only attributes your model-token spend.

Frozen cut: 0.5.2, protocol v14, Linux. Pair it with CLI 0.41.5 (this checkout or the GitHub release). Tag sdk-v0.5.2 — not v0.5.2, which is a historical CLI tag.

python3 -m pip install \
  "https://github.com/OpenPeach-ai/dgc/releases/download/sdk-v0.5.2/dgc_sdk-0.5.2-py3-none-any.whl"

Or from a clone: python3 -m pip install -e sdk/python.

That URL installs our dgc-sdk wheel (import dgc_sdk). Do not run pip install dgc: PyPI’s package named dgc is a different project. This SDK is not on the PyPI or npm indexes yet. The wheel still needs a runtime that can run python -m dgc serve (this repository or CLI 0.41.5). Set DGC_PYTHON if python3 cannot import dgc.

Each DGC owns one state_dir HOME. Host ~/.dgc is unused unless you pass inherit_user_state=True — do not, in production embeds.

from dgc_sdk import DGC, Pricing, RuntimePolicy

dgc = DGC(
    state_dir="/var/lib/myapp/dgc",
    inherit_user_state=False,
    department="platform",
    pricing=Pricing(input_per_million=0.0, output_per_million=0.0),
    policy=RuntimePolicy(network="deny", deny_tools=("write_file",)),
)
session = dgc.session(cwd="/path/to/workspace", permissions={"mode": "default", "unhandled": "deny"})
result = session.run("Summarize this repository. Do not edit files.")
print(result.status, result.usage)

RuntimePolicy(deny_tools=("write_file",)) is compiled into isolated deny rules in every permission mode, including auto. Write-tool denies also block bash file-writes (redirects, tee, cp/mv).

TypeScript lives in sdk/typescript (Node ≥ 22). Usage, audit, and retry helpers are Python-first.

Source: github.com/OpenPeach-ai/dgc (sdk/).