Modal
Serverless sandboxed runtime with gVisor-isolated containers for agent code execution.
Maturity tier details
Computed tier: 3(raw score 3). Effective tier: 3.
Facets:documented_in=0evaluated_on=0integration_surfaces=1paper_citations=0recent_commit=2stars=0
Source of truth:pipeline/maturity/score.py
Description
Serverless compute platform whose Sandbox API exposes per-call, gVisor- isolated containers designed for executing AI-generated code. Supervision primitives: ephemeral container lifecycle, fine-grained CPU/GPU/memory limits, filesystem and network policy controls, and snapshot/restore for state checkpointing. Tuned for fully_autonomous agent back-ends that need to run untrusted code at low startup cost.
Why it's included
Occupies recovery × post_generation × fully_autonomous alongside E2B and Daytona. Included because gVisor-based isolation and snapshot/restore represent a distinct point in the sandbox-provider design space relevant to autonomous coding-agent deployments.
Security notes
Proprietary hosted platform. Sandboxes run on gVisor for kernel-surface reduction; still evaluate tenancy, data-residency, and egress-control guarantees for sensitive workloads. Agent-generated code should be treated as untrusted — pin network policy and secret scopes explicitly per sandbox.
Addresses failure modes
- Supply-chain attack — Slopsquatting, typosquatting, compromised MCP servers.
- Security vulnerability — Insecure patterns, injection flaws, weak cryptography, secret leakage.
- Logic error — Correct API use, incorrect logic.