SecureVector SDK for CrewAI

Bring the SecureVector local threat monitor's three controls — tool-call permissions, secret / data-leak detection, and threat detection — to every CrewAI tool call, with tamper-evident audit logging.

pip install securevector-sdk-crewai

📦 One install — batteries included. pip install securevector-sdk-crewai also installs the local SecureVector app (securevector-ai-monitor): the adapter and the detection engine + tamper-evident audit chain arrive in a single pip install. The SDK is a thin interception layer — the app must be running locally (securevector-app --web) for it to do anything.

🌐 Pointing at your own cloud? Use the lightweight install. If you've deployed SecureVector to your own cloud, you don't need the bundled local app. Install only the adapter on the machine where your agents run, and point it at your deployment:

# lightweight — adapter only, no local app (your env already has crewai)
pip install securevector-sdk-crewai --no-deps

# point at your SecureVector endpoint — all you need for a private (in-VPC) endpoint
export SECUREVECTOR_SDK_APP_URL=https://<your-securevector-endpoint>

# OPTIONAL: only if your endpoint is publicly exposed and gated with an inbound token.
# A private endpoint in your own VPC needs no key. To gate a public one, use a free
# SecureVector cloud account API key or an SVET token — it gates access only; no agent
# data is sent to SecureVector.
export SECUREVECTOR_API_KEY=<SecureVector account key or SVET token>

The adapter then forwards every tool call to your remote deployment instead of a local app. The default pip install securevector-sdk-crewai (no --no-deps) still bundles the app for local use.

Quick start

Wrap your tools (recommended — robust across CrewAI versions):

from securevector_sdk_crewai import secure_tools

agent = Agent(tools=secure_tools(my_tools), ...)

or install globally (best-effort monkeypatch of CrewAI's BaseTool):

from securevector_sdk_crewai import install

install(mode="observe")

or fully zero-config:

import securevector_sdk_crewai.auto   # reads env, installs globally

LLM cost tracking — install() / .auto also patch Crew.kickoff so every run's token usage (prompt / completion / cached) posts to the app's Cost Tracking, where the pricing table turns it into real dollars (your crew runs on your own API keys) and per-agent budgets apply. If you only wrap tools with secure_tools(), post usage explicitly after a run:

from securevector_sdk_crewai import track_crew_usage

result = crew.kickoff()
track_crew_usage(crew, agent_id="research-crew")   # agent_id optional

Records are attributed to agent_id="crewai-agent" by default — override via track_crew_usage(crew, agent_id=...) or SECUREVECTOR_SDK_AGENT_ID. Cost capture is best-effort: an unreachable app never breaks the crew.

What happens on every tool call

Before a tool runs, the SDK:

1. (a) Permissions — resolves an allow/block verdict for the tool, using the app's own precedence: cloud-pushed synced policy → local override → essential registry → default-allow.
2. (b)+(c) Secret & threat scan — sends the serialized tool input through the app's /analyze pipeline.

After the tool returns, the result is scanned the same way to catch secrets / exfiltration in tool output. Every decision is written to the app's audit chain tagged runtime_kind="crewai".

observe vs enforce

install(mode="enforce")   # blocks denied tools and fails closed if the app is down

In enforce mode a denied tool raises ToolBlocked before it executes; enforce also prints a one-time disclosure to stderr.

Configuration

All optional, via env or kwargs:

Compliance

The tool-call-level, attributed, tamper-evident audit trail this produces is exactly the action-layer logging auditors ask for under EU AI Act Art. 12 / 15. This SDK produces the local evidence; the cloud governance surface turns it into an auditor-ready pack.

Trademarks

SecureVector is the product name of this SDK. CrewAI is a trademark of CrewAI, Inc. This is an independent, community SDK that integrates with CrewAI via its public tool API. It is not affiliated with, sponsored by, or endorsed by CrewAI, Inc. The name uses "crewai" only descriptively, to identify the framework this package works with (nominative fair use).

License

Apache-2.0. See LICENSE and NOTICE.