Blacksand
Open-source · AI-assisted security research
arzumanabbasov/blacksand
Blacksand is a self-hosted workspace for AI-assisted security research. It turns a research objective into a visible workflow where a planner creates work intents, concurrent workers execute them, and findings, evidence, conversations and decisions remain together.
Why it exists
Security research agents need persistent task context and reviewable evidence across concurrent work. A generated report alone is not enough: people need to inspect the supporting evidence, retest findings, control execution and understand what happened during a run.
How it works
The workspace connects shared asset scope with task goals, intents, hints and findings. It supports local tools or remote execution through E2B, selectable model providers, MCP clients, approvals and interception rules. Reports can include HTTP evidence bindings, retests and exports.
Usage & impact
An actively developed, self-hosted open-source project with setup and architecture guides, automated CI and a public field journal for real runs and experiments. Use it only in authorized environments and review the evidence: model output is not an independent vulnerability verdict.