Govern coding-agent changes in OpenAI Codex with signetry-core.
pip install "signetry-core @ git+https://github.com/Signetry/core@v0.8.0"Add a .signetry/admission.yaml to your repo (allowed/forbidden paths, diff budget,
required checks). A conservative default applies without one.
Codex reads MCP servers from ~/.codex/config.toml. Add Signetry's server so the
agent can run admission / verify / provenance itself:
[mcp_servers.signetry]
command = "python"
args = ["-m", "signetry_core.mcp_server"]
[mcp_servers.signetry.env]
SIGNETRY_MCP_ROOTS = "/absolute/path/to/your/repo"SIGNETRY_MCP_ROOTS scopes the server to your workspace(s) so it can't be pointed
at arbitrary host paths. The agent then has signetry_admit, signetry_verify, and
signetry_provenance tools.
Codex supports lifecycle hooks. Configure a hook that runs signetry guard before a
file write / command, so an out-of-scope or forbidden action is blocked by
deterministic code (not the model). See the Codex config docs for the exact hook
schema for your version; the guard command to wire in is:
signetry guard --repo "$REPO" --path "$PROPOSED_PATH" # exit 1 = deny
signetry guard --repo "$REPO" --command "$PROPOSED_COMMAND" # exit 1 = denysignetry guard exits non-zero and prints a reason when the action violates the
contract; exit 0 means allowed.
Whichever agent opens the PR, make Signetry Admission a required check so nothing merges without a signed receipt: https://github.com/marketplace/actions/signetry-admission. In-editor guards are best-effort defense-in-depth; the CI check is the enforced gate.
signetry-core also finds vulnerabilities and can govern the fix with Codex:
signetry scan . # SAST over the repo (7 languages, SARIF), offline & free
signetry scan . --fix --fix-agent codex-cli # draft a governed fix → admission → signed receipt--fix is bring-your-own-key (your OPENAI_API_KEY, never shared, redacted
from every artifact) and opens branch-only PRs — never merges. Works with
OpenAI-compatible gateways (e.g. IBM ICA) via --codex_model / base-URL inputs.
See signetry-core: AUTOFIX_SETUP.md.
Part of the Signetry platform. Governance logic lives in signetry-core; this integration never reimplements policy and never auto-merges.
Apache-2.0. Use it, fork it, ship it commercially — no strings.
This repository is part of Signetry's open-core model:
the integration surface is Apache-2.0 so anyone can add an agent, an editor, or a
CI adapter, while the engine (Signetry/core) is
source-available under BUSL-1.1 and converts to Apache-2.0 on 2030-08-31.
Contributions are accepted under the CLA — it lets us move a well-built adapter into the engine later without asking every contributor for permission again.