Execution sandbox for AI agents with three-stage sanitize, adjudicate, execute pipeline
sandraschi/deepfang is an MCP server distributed on github, maintained by sandraschi, tracked here since June 2026. It has shipped 0 releases and exposes 7 tools. Tools include deepfang_adjudicate, deepfang_agentic_workflow, deepfang_audit, deepfang_dispatch, deepfang_pipeline, deepfang_sanitize, and 1 more. Its composite risk grade is A — an inferred review prompt computed from observed signals, not a verdict.
-
deepfang_adjudicateStage 2 only — LLM verdict with rationale -
deepfang_agentic_workflowMulti-step goal via FastMCP 3.2 sampling -
deepfang_auditQuery the adjudication log -
deepfang_dispatchStage 3 only — requires prior approval in audit log -
deepfang_pipelineFull pipeline in one call — the main entry point -
deepfang_sanitizeStage 1 only — fast regex scan, returns threat score -
deepfang_statusHealth check across all pipeline services
No releases yet.
No releases yet.