Skip to main content

Local-first AWS cost analysis and evidence CLI. Read-only by construction.

Project description

Kulshan

Local-first AWS cost analysis and evidence CLI.

Something changed in AWS. Kulshan helps you investigate what moved, what evidence supports the explanation, what evidence is missing, and how confident the conclusion should be.

pip install kulshan
aws login
kulshan preflight
kulshan report

PyPI License Python


What you get

Kulshan produces a local HTML and JSON report containing:

  • Executive cost movement summary
  • Supporting evidence (anomaly detection, usage-type attribution, period deltas)
  • Contradicting or incomplete evidence (missing tags, gaps in coverage)
  • Ownership and attribution confidence
  • Coverage disclosure and unknowns
  • Recommended next investigation steps

A check is marked clean only when the required AWS evidence was successfully retrieved and evaluated. Failed evaluations are reported as "could not check" with the denied action named, never silently passed.

View the synthetic sample report

Kulshan report preview showing cost investigation structure: question, conclusion, supporting evidence, contradicting evidence, ownership confidence, and next step

Local-first. Read-only. No SaaS. No telemetry.


How it works

kulshan report                                    # cost investigation (default)
kulshan report --packs security,sweep             # add security + waste detection
kulshan report --packs all --regions us-east-1    # full diagnostic across all packs
kulshan analyze cost --path ./cur/ --month 2026-06  # investigate CUR data locally

Reads Cost Explorer data and your own CUR/Data Export Parquet files in place. No data leaves your machine. DuckDB queries locally. No Athena, no Glue, no data warehouse.


Trust model

Read-only by construction, not by default. There is no cleanup mode, no write path, no telemetry to opt out of.

Documentation | IAM Policy | Changelog


Install

pip install kulshan

Python 3.9+. macOS, Linux, Windows. Optional extras: kulshan[pdf], kulshan[excel], kulshan[pptx], kulshan[mcp], or kulshan[all].


Credentials

If aws sts get-caller-identity works, Kulshan works.

aws login
kulshan preflight
kulshan report

Named profiles, environment variables, and role assumption all work:

kulshan --profile production report
kulshan --role-arn arn:aws:iam::123456789012:role/KulshanAudit report

Run kulshan preflight to check connectivity and permissions before scanning.


CUR / Data Export Investigation

Query your CUR Parquet files locally or from S3. No Athena, no Glue, no data warehouse. DuckDB queries in place.

kulshan cur validate --path ./cur/
kulshan analyze cost --path ./cur/ --month 2024-06
kulshan analyze ec2 --cur ./cur/ --month 2024-06
kulshan analyze cost --s3 s3://bucket/prefix/ --month 2024-06

Top movers by service, account, region, usage type. Period-over-period deltas. Resource-level contributors. Tag coverage. All outputs include provenance, evidence IDs, and human_review_required: true.


The 10 Audit Packs

Pack What it watches
cost Anomalies (z-score, IQR, MAD), commitment gaps, spend acceleration, forecasts
security IAM, encryption, network exposure, logging, public access, GuardDuty
sweep Orphaned volumes, unused EIPs, idle LBs, detached ENIs, empty repos
dr Backup coverage, single-AZ, single points of failure, missing replication
age EOL runtimes, expiring certs, stale AMIs, outdated engines
drift CloudFormation drift, IaC coverage, severity classification
tag Missing required tags, unattributed spend, key inconsistencies
pulse Alarm gaps, missing metric filters, blind spots
limit Quota headroom, at-limit services, scaling risk
topo CIDR overlaps, route integrity, peering issues, TGW misconfigs
kulshan report                                    # cost only (default, ~$0.15)
kulshan report --packs security,sweep             # specific packs (free APIs)
kulshan report --packs all --regions us-east-1    # full diagnostic

Automatic Environment Isolation

On first run, Kulshan identifies your AWS principal, creates an isolated local environment, and routes all data there. Different identities get separate environments. No flags required.

✓ Created environment readonlyrole-cedar
  Using readonlyrole-cedar · account 1234…5678

When CUR data reveals a payer account, the environment binds to that payer. Multiple identities accessing the same payer can be reconciled into a unified timeline:

kulshan workspace reconcile

Workspaces with multiple connections produce consolidated reports automatically - one scan, all connections, deduplicated findings, per-connection coverage metadata.


MCP Server

Kulshan exposes its findings to MCP-compatible agents (Claude Desktop, Cursor, Kiro, others). Deterministic evidence in, agent reasoning out.

kulshan mcp-serve
{
  "mcpServers": {
    "kulshan": { "command": "kulshan", "args": ["mcp-serve"] }
  }
}

Seven tools: kulshan_preflight, kulshan_report, kulshan_quick_security, kulshan_list_packs, kulshan_cur_validate, kulshan_analyze_ec2, kulshan_analyze_cost.


Output Formats

kulshan report -o report.html           # Self-contained HTML report
kulshan report --format json -o s.json  # Structured, machine-readable
kulshan report --format sarif -o r.sarif # GitHub Security tab
kulshan report --format csv -o f.csv    # Spreadsheet / JIRA import
kulshan convert -i scan.json -o r.html  # Re-render without re-scanning

Account IDs redacted by default. --show-pii for full IDs. Atomic writes prevent partial files.


CI/CD

kulshan report --packs security --format sarif -o results.sarif --yes --no-history

Exit code 1 when critical findings are present - use as a quality gate. SARIF uploads to GitHub Code Scanning. Full GitHub Actions and GitLab CI examples in docs/ci-cd.md.


Quick Reference

kulshan --version                       # Version
kulshan preflight                          # Check credentials and permissions
kulshan report                          # Cost baseline (default)
kulshan report --quick                  # Skip confirmation
kulshan report -o report.html           # HTML report
kulshan report --packs all --regions us-east-1 --deep  # Full deep scan
kulshan report --perf                   # Show API timing
kulshan history                         # Past scans
kulshan history --direct-only           # Current workspace only
kulshan workspace list                  # All environments
kulshan workspace reconcile             # Link shared-payer environments
kulshan shell                           # Interactive REPL
kulshan convert -i scan.json -o r.html  # Re-render

AWS API Cost

Cost pack: ~$0.15 (CE API at $0.01/request). All other packs use free APIs. Kulshan confirms before making CE calls. Use --yes in CI/CD.


About the Name

Kulshan is the Lummi name for the mountain known colonially as Mt. Baker, meaning "great white watcher." The mountain is visible from Mission, BC and is an active volcano in the Cascade Range. We acknowledge the Lummi and Nooksack peoples as the original namers of this mountain.


Maintained by

Mission FinOps - open-source AWS audit tooling.


License

Apache 2.0. Free and open source forever.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

kulshan-0.4.4.tar.gz (323.6 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

kulshan-0.4.4-py3-none-any.whl (415.2 kB view details)

Uploaded Python 3

File details

Details for the file kulshan-0.4.4.tar.gz.

File metadata

  • Download URL: kulshan-0.4.4.tar.gz
  • Upload date:
  • Size: 323.6 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.14

File hashes

Hashes for kulshan-0.4.4.tar.gz
Algorithm Hash digest
SHA256 094ccc66d0a8b231bd3fe0fd5b2433d744a9d7ed1d837ae6acabefe780041ac0
MD5 a1de01e8161c8bb78e3b9ecf65c4412a
BLAKE2b-256 c33e5e5ab1a095fb5b1b50446172ff7fe715d60aa3a8d9a2c0fa19343b284d05

See more details on using hashes here.

Provenance

The following attestation bundles were made for kulshan-0.4.4.tar.gz:

Publisher: publish.yml on MissionFinOps/kulshan

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file kulshan-0.4.4-py3-none-any.whl.

File metadata

  • Download URL: kulshan-0.4.4-py3-none-any.whl
  • Upload date:
  • Size: 415.2 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.14

File hashes

Hashes for kulshan-0.4.4-py3-none-any.whl
Algorithm Hash digest
SHA256 fc8be0e08f8f5dbfb379f267ea839b20128b4bf48c2267f016063d6fc9124af5
MD5 6b0952a0518086cbca6da71fa73a96a4
BLAKE2b-256 16917a2a032234afc6b0d264c732a34db2e5095728e977dfdcaa033978845dbd

See more details on using hashes here.

Provenance

The following attestation bundles were made for kulshan-0.4.4-py3-none-any.whl:

Publisher: publish.yml on MissionFinOps/kulshan

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page