SigComply runs as a step in your GitHub Actions or GitLab CI pipeline. Once a day it reads your cloud and SaaS settings, checks that your policy documents are on file, and signs the evidence into storage you own. A failing policy fails the build.
It covers SOC 2 (2017 Trust Services Criteria) and ISO/IEC 27001:2022. The evidence goes to an S3, GCS or Azure Blob bucket, or a local folder. Pull requests and pushes to main rerun the setting checks. If you connect the repository to a project, the run also sends pass/fail counts per policy to the SigComply dashboard. Otherwise nothing is sent.
Your systems
AWS, GCP, Azure, GitHub, GitLab, Okta, Active Directory APIs
Your documents folder (policies, training records, reviews)
sigcomply check, in your CI
Reads with the pipeline's credentials, evaluates policies, signs the evidence, sets the exit code
Your bucket
Signed evidence and results
SigComply dashboard (optional)
Counts per policy only
The CLI reads .sigcomply.yaml, works out the framework and the audit period (for example 2026-Q4), and matches each policy to the sources that provide its evidence.
Each source reads its API, or the manual-evidence folder, and turns the response into evidence records. The CLI checks each record against its schema, signs it, and writes it to your storage.
Each policy evaluates its records and passes, fails or skips. A policy with no configured source is skipped and left out of the score.
The CLI writes the results and a signed manifest of every file the run wrote.
A failing policy exits 1 and fails your build; skipped policies don't fail the run. Exit 2 means an execution error and 3 a config error.
The CLI sends the per-policy counts to the SigComply dashboard only when you set cloud.base_url, and only from a full run in CI such as the daily scheduled run.
Automated policies read your settings through each provider's API (AWS, GCP, Azure, GitHub, GitLab, Okta and Active Directory): whether every user has MFA, whether storage is encrypted, whether main needs a review. They are the only checks that judge a control.
The credentials stay in your CI. On GitHub Actions and GitLab CI the workflow that
sigcomply init-ci writes assumes an AWS role over OIDC,
so you store no long-lived AWS key. The role reads your systems and writes only to your
evidence bucket. SaaS tokens live in your CI secrets. No vendor role is created in your
account.
Your storage keeps the records each policy evaluated, in a normalized form. Every policy is code in the CLI's public source, so you and your auditor can read exactly what it checks. You can add your own policies in YAML or Rego in your repository; the CLI loads them on every run.
Some controls no API can answer: a security policy, a training record, an access review.
For these, you put the document in a folder in your storage, one folder per catalog entry
and period, for example manual/access_review_quarterly/2026-Q4/.
sigcomply evidence due lists the folders still empty.
The CLI checks that:
It merges the files into one PDF, records each file's hash, signs that record and stores the merged PDF with the run.
The CLI never reads or judges what a document says. A manual policy passes when the document is on file. Your auditor reads what it says.
For declarations and checklists, the Evidence forms page turns a form into a PDF in your browser. It uploads nothing; you put the PDF in the folder.
Every evidence file is signed with Ed25519, using a new key for each signature; the public key and the signature are stored in the file. Every run also writes a manifest that lists the SHA-256 hash of each file it wrote, signed the same way.
To verify a whole run, run
sigcomply report --period 2026-Q4 --view integrity. It
re-hashes every file and checks the manifest's signature.
To check one file, drop it into the Evidence Verifier, with its merged PDF for a document. It runs in your browser and sends nothing anywhere.
A file changed or removed after the run fails verification, and so does a file added to the run folder later.
It doesn't prove who signed it. Someone with write access to your bucket could replace a file and sign it again. Keep the bucket write-once, with S3 Object Lock or GCS Bucket Lock, so nobody can. The CLI doesn't turn that on for you.
What leaves your pipeline: pass/fail counts per policy and run details such as the framework, period, repository, branch, commit, CLI version and CI run link. Never evidence files or their hashes, and never the names, emails or cloud account IDs of the resources and people checked.
Run sigcomply check --capture-cloud-payload payload.json to
write the exact payload to a file and read it before you turn submission on.
The CLI does everything above for free. SigComply Cloud (paid, with a 2-month free trial) adds what one run can't:
The reports are a compliance summary, a drift report and a control matrix, for any date range, and each auditor grant can have an expiry date.
Your runs are always recorded. Without a plan, the dashboard shows each project's latest status and the last 30 days of runs; older runs come back when you upgrade.
See the Overview, a project's status page and the Audit window with example data.
The dashboard works only from the counts above. Your auditor checks the evidence itself in your
storage. The ISO 27001 Statement of Applicability comes from the free CLI
(sigcomply report --view soa).
SigComply Cloud is $39/month.
Start the free trialSigComply doesn't run vendor reviews, security training, device monitoring, background checks, a trust center or your audit. A GRC platform covers more of that, and needs more access to do it.
SigComply collects evidence and evaluates policies. Your auditor decides whether you are compliant.
The Quickstart runs SigComply with manual evidence only, so it needs no cloud credentials. You make one policy pass, verify the signatures, and see verification fail after you change a file.
$ sigcomply init$ sigcomply init-ci --ci github
Run SigComply in CI covers the rest step by step: a bucket, a CI role over OIDC, the first run and your documents.
The CLI's docs have every detail. The architecture docs are written for contributors.