redcell
AI red-team platform. Autonomous LLM agents run a penetration test end to end inside a Kali container and write the report. LangGraph plan/act engine, provider-
What is redcell?
AI red-team platform. Autonomous LLM agents run a penetration test end to end inside a Kali container and write the report. LangGraph plan/act engine, provider-agnostic models via LiteLLM, PDF/JSON/SARIF output. FastAPI + React.
Overview of redcell
What it is
REDCELL is an AI red-team platform that runs autonomous LLM agents to perform a penetration test end to end inside a Kali container and write the report. It uses a LangGraph plan/act engine and is provider-agnostic via LiteLLM, supporting models from OpenAI, Anthropic, Google, GLM, DeepSeek, Kimi, local Ollama, and others. Output formats include PDF, JSON, and SARIF. The platform is built with FastAPI and React.
Key features
- Multi-agent engine with a LangGraph plan/act loop: an orchestrator delegates objectives to executor agents that run shell tools and record findings, loot, and hosts.
- Structured tools for nmap, nuclei, directory and vhost discovery, and Metasploit search and run; each parses its own output into the attack surface or findings.
- Real execution in a Kali container over
docker exec, supporting local or remote servers over SSH with host networking. - Agent browser: a real Chromium in the Kali container for login flows and JavaScript-heavy apps, viewable live and controllable by the operator.
- Reverse shells: an agent opens a listener, catches the shell, and provides an interactive terminal; network pivoting via chisel reverse SOCKS tunnel.
- Chat-driven operation: the chat steers the live orchestrator, can reopen finished runs for new work, and answers questions about the engagement.
- Code-scan sessions for source-code security review of public git repos or local folders, with findings mapped to file and line.
- Per-session configuration for execution server, model, and optional egress proxy, with a connection test before use.
- Findings triage: verify or dismiss findings and merge duplicates; the report excludes dismissed ones and marks verified ones.
- Reports: export PDF, JSON, and SARIF; the write-up includes executive summary, methodology, findings, and remediation in priority order.
- Live console with activity feed, terminals, agent browser streaming over WebSockets, and agent graph, findings, loot, attack surface, listeners, and proxy history updating on a short poll.
- Notifications: in-app toasts and browser notifications when the tab is in the background.
Use cases
- Authorized penetration testing of systems you own or have explicit written permission to test.
- Security research and education in a controlled environment.
- Automated security assessment of web applications, networks, and source code.
- Red team exercises where autonomous agents simulate attacker behavior.
Getting started
You will need Docker, uv, and bun.bash# 1. infrastructure (Postgres, Redis, MinIO)docker compose -f docker-compose.dev.yml up -d# 2. Python deps, database, and seed datauv sync --group liveuv run rc db upgradeuv run rc seed # admin user, provider catalog, buckets# 3. copy the env templatecp .env.example .env# 4. run the three processes (separate terminals)cd apps/api && uv run uvicorn app.main:app --host 127.0.0.1 --port 8080cd apps/worker && uv run arq worker.settings.WorkerSettingscd apps/web && bun install && bun run devOr start all three at once with a process manager: pipx install honcho then honcho start.Open http://localhost:5183 and sign in with admin / admin.
License and pricing
License: MIT. Pricing: open source.
Based on the project's public README and repository metadata.