Updated Agent Sandbox Runtime and Upgrade Guidance Released for Self-Hosted AI Agent Platforms

The guidance covers revised environment configurations, deprecated settings, mandatory image updates, and operational safeguards for Agent Sandbox-enabled deployments

HANGZHOU, China, Sept. 14, 2026 /PRNewswire/ -- FastGPT has published consolidated Agent Sandbox runtime and deployment guidance covering changes introduced from v4.16.0 onward. Sandbox-reliant deployments must update the Agent-sandbox-proxy service, with optional PORT and PREVIEW_PORT configuration to split WebSocket and HTTP traffic if gateways lack unified port support. An https://{{host}}/health endpoint validates proxy connectivity, and image tags differ by component: fastgpt-app and fastgpt-pro move to v4.16.0, fastgpt-plugin to v1.1.0, and agent-sandbox-volume and agent-sandbox-proxy to v0.3.0.

Administrators must adjust environment variables for fastgpt-app and fastgpt-pro services, adding AGENT_SANDBOX_PREVIEW_PROXY_URL plus the OpenSandbox volume-name prefix, listed as VM_VOLUME_NAME_PREFIX in the v4.16.0 upgrade note and as AGENT_SANDBOX_OPENSANDBOX_VOLUME_NAME_PREFIX in current environment-variable documentation; confirm the name against the documentation for the version being deployed. Deprecated variables include AGENT_SANDBOX_DISK_MB and all E2B-related configurations; existing E2B users must switch to opensandbox or sealosdevbox providers and remove AGENT_SANDBOX_E2B_API_KEY. New optional sandbox variables allow tuning of CPU count (default 1), memory limit (default 2048 MiB), storage capacity (default 1 GiB), auto-suspend time (default 60 minutes), and auto-archive time (default 7 days) with documented default values.

Mixed version deployments of FastGPT, sandbox proxy, and sandbox components are not supported. The platform does not enforce cross-origin isolation for sandbox proxy deployments, though cross-origin configuration is strongly recommended to mitigate security risks. Unsupported use cases include legacy E2B setups, non-standard container orchestration topologies, and custom sandbox provider integrations outside officially supported options. Deployment-specific network rules or resource constraints may alter default behavior.

About FastGPT

FastGPT is an open-source AI application platform offering RAG knowledge bases, visual workflows, agent orchestration, Skill, MCP and multi-channel publishing, available as a cloud service, community self-hosted, or commercial private deployment. As of Sept. 7, 2026, the GitHub repository labring/FastGPT has 29,592 stars and 7,301 forks across 275 releases, with v4.16.2 published on Sept. 3, 2026. Repository: github.com/labring/FastGPT

SOURCE FastGPT