SandBase Harness
Local-first Apache-2.0 agent runtime with sandboxed sessions, memory, credentials, MCP toolsets, audit/replay and a local Console.
Tags
Description
Open-source self-hosted runtime (sandbaseai/sandbase-harness; not archived; Apache-2.0). README and llms.txt describe persistent sessions, credential vaults, permission policies, MCP toolsets and sandbox backends (local process, Docker, Kubernetes, self-hosted workers). Latest tagged release cited in the project's llms.txt: v0.3.8 (verified 2026-11-10). Local-first SQLite/file storage; no hosted control plane is required. Also referred to in-repo as managed-agents; README warns to install only from official paths once a scoped package is announced. Sandboxes can run generated code and reach external services — isolation and permissions are the operator's responsibility. Model-provider credentials and compute are separate. Same operator family as the hosted SandBase service and sandbase-skills (AXIOMCLOUD / sandbaseai).
Agent access
Related
- SandBase: Hosted one-key API for models, data APIs, agent sandboxes and MCP, with pay-per-call billing and a free catalog/discovery surface.
- SandBase Skills: Apache-2.0 collection of installable Agent Skills for research, marketing, engineering and business workflows, maintained by SandBase.
- E2B: Open-source, secure cloud sandboxes for AI agents: an isolated machine per agent to run code, browse and use tools.
- OpenAgentCore: MiniMax's open-source, self-hosted implementation of OpenAI's Agents API that runs Codex, Claude Code or MiniMax Code sessions in sandboxes.
Sources
- https://github.com/sandbaseai/sandbase-harness
- https://github.com/sandbaseai/sandbase-harness/blob/main/LICENSE
- https://github.com/sandbaseai/sandbase-harness/blob/main/llms.txt
- https://github.com/sandbaseai/sandbase-harness/blob/main/llms-install.md
- https://github.com/sandbaseai/sandbase-harness/blob/main/docs/README.md
Machine-readable
- JSON
- Markdown
- Source file on GitHub (edit via pull request)