VELA
Vela delivers isolated execution for untrusted code in AI agent workflows.
Open SourceCoding & Dev
Visit website Free to browse · updated 2026-06-18
What is VELA?
Vela addresses risks associated with executing code generated by AI models. By leveraging dedicated virtual machines for each task, it prevents direct access to the underlying system and supports defined limits on operations. The system includes mechanisms for authorizing specific capabilities and recording execution details in structured formats. These elements work together to enable monitoring and policy enforcement without relying on traditional container approaches. Integration options exist for various AI development frameworks, and the project is distributed under an open source license that permits broad use and modification.
Key features
Firecracker micro-VM isolation
HMAC capability tokens
Full audit trail
Policy engine with YAML deny lists
Intrusion detection
Prometheus metrics
Framework adapters for LangChain, LlamaIndex, CrewAI, OpenAI
What you can use VELA for
Secure AI Agent Code Execution
Isolate and run untrusted code generated by AI agents inside Firecracker micro-VMs, preventing jailbreaks or malicious prompts from affecting the host system.
Policy-Controlled SaaS Workloads
Apply HMAC-signed capability tokens and YAML deny lists to restrict filesystem access, network calls, and syscalls for code running on behalf of users or models.
Audited Execution Environments
Capture every execution event in append-only JSONL logs with real-time streaming and Prometheus metrics for compliance and incident response.
How to use VELA
- 1Install the Rust daemon with cargo install vela-engine
- 2Install the Python client with pip install vela-agent
- 3Generate an HMAC capability token
- 4Start the Vela daemon and load policy.yaml
- 5Import AegisPythonREPL or framework adapter and run code
VELA pricing
Pricing model: Open Source. Plan details are indicative — check the site for current prices.
Self-hosting
Free
- MIT License
- Free forever for self-hosting
- Full source access
Cloud
PopularCustom
- Managed infrastructure
- Waitlist only
Editor's verdict
Pros
- +Open source under MIT License
- +Fast ~150ms p50 with pre-warmed pool
- +Easy integration with AI frameworks
Cons
- –No public pricing information
- –Cloud version on waitlist
- –Requires self-hosting setup
Our take: VELA is a solid coding & dev choice. It's valued for open source under mit license and fast ~150ms p50 with pre-warmed pool. The main trade-off is no public pricing information. A good pick if you want capable AI without a high upfront cost.
Frequently asked questions
Vela boots each execution in a fresh Firecracker micro-VM from a pre-warmed pool, providing hardware-level kernel isolation with no shared host kernel.
Summary
VELA is a solid coding & dev choice. It's valued for open source under mit license and fast ~150ms p50 with pre-warmed pool. The main trade-off is no public pricing information. A good pick if you want capable AI without a high upfront cost.
Did you find this helpful?
User reviews
Verified reviews from the community shape this tool's rating.
Loading reviews…
VELA alternatives
Similar coding & dev tools worth comparing.
