The AI agent that orchestrates your development team
Argy Code turns one engineer into a governed team. Security audits, development, testing, documentation and research run in parallel — under your enterprise compliance controls.
Useful links: Argy Code documentation · multi-provider LLM Gateway · enterprise RAG memory · AI engineering platform · AI security governance
Complex projects that can run for days — safely
The orchestrator plans, delegates, validates and resumes work across sessions — with persistent checkpoints, progress summaries and human approvals before risky decisions. Multi-day migrations, refactors, audits, and cross-team releases continue without losing context. Ideal for legacy system modernizations (SAP, COBOL, mainframes, ERP) and critical multi-team migrations spanning several weeks.
Orchestrator
Mission queue, routing, checkpoints, final synthesis
Builder
Implementation, TDD, refactoring
Planner
Architecture, planning, ADRs
Explorer
Codebase mapping and impact analysis
Code-Reviewer
Quality gates, patterns, security review
Sec-Auditor
SAST, SCA, IaC, secrets, DAST
Doc-Generator
.pptx, .xlsx, .pdf, .docx outputs
Browser
Navigation, E2E, WCAG, Electron apps
Deep-Researcher
Multi-source research and cited synthesis
Skill-Creator
Reusable SKILL.md modules
General
General-purpose execution for delegated scopes
Token optimization & checkpoint/resume
Argy Code automatically compacts context at 80% of the window, preserves critical session context, and applies a multi-layer governance optimization. If a session is interrupted, checkpoint/resume recovers partial work instead of starting over.
Built-in capabilities
A complete development agent: from code and security to documents, browser workflows, and research.
Browser automation
Full web navigation, E2E tests, WCAG/ARIA accessibility audits, scraping, Electron app automation (VS Code, Figma, Slack), HAR capture, Web Vitals, and React profiling.
Automated security auditing
8 phases: SAST (Semgrep), SCA (Trivy), IaC, Secrets (gitleaks), DAST (Nuclei). OWASP Top 10, PCI-DSS 4.0, SOC2, ISO 27001. Reports in .argy/audits/ — HTML + Markdown.
Document generation
PowerPoint (.pptx), Excel (.xlsx), PDF (.pdf), and Word (.docx) reports with enterprise visual charters. Workflow: generate TypeScript → execute via bun → deliver the file.
LSP — code intelligence by default
Language Server Protocol enabled by default: real-time diagnostics, symbol search, go-to-definition, find references, and semantic refactoring. No experimental toggle required.
Enterprise multi-scope RAG
Persistent enterprise RAG memory across USER · PROJECT · ORG · TENANT. Agents query memory before acting to follow established patterns.
ACP + MCP
ACP (Agent Client Protocol) powers VSCode and JetBrains integrations. MCP (Model Context Protocol) connects Argy Code to external tool servers: databases, APIs, and DevOps tools.
Skills & reusable modules
SKILL.md modules injected on demand. Automatic discovery (.argy/skill/, ~/.argy/skill/), per-skill permissions (ask/allow/deny), and a Skill-Creator agent to generate new skills.
Deep research
Deep-Researcher runs structured multi-source investigations (web, RAG, parallel sub-agents) with a 3-phase pipeline: Outline → Investigation → cited synthesis.
Governed AI access
Every request goes through the multi-provider LLM Gateway: quotas, PII/secrets/prompt-injection filters, risk assessment before sensitive actions, rate limits, and full auditability.
8 security audit phases in one command
Sec-Auditor orchestrates a complete audit aligned with OWASP Top 10, OWASP API, OWASP LLM 2025, PCI-DSS 4.0, SOC2, ISO 27001, and GDPR. CycloneDX SBOM and CVE scoring included. See AI security governance. Run a complete security audit with a single command.
Reconnaissance
Map attack surface and assets
SAST
Semgrep static analysis, enterprise rules
SCA
Dependency vulnerabilities with Trivy
IaC
Terraform / Kubernetes misconfiguration
Secrets
gitleaks detection — no cleartext secrets
DAST
Nuclei with explicit consent gate
Triage
CVE scoring, OWASP, reachability
Report
HTML + Markdown in .argy/audits/
Built for CIOs, CTOs, and platform teams
Argy Code is part of a complete AI engineering platform: project doctrine, tenant isolation, audit trail on every action, and a configurable policy engine.
Project doctrine file (committed to Git)
A doctrine file committed to git. Team conventions, architecture decisions, constraints, and workflow rules. Kept in sync with enterprise RAG memory.
Multi-tenant isolation
Every data query scoped by tenantId. Permission inheritance without privilege escalation. Signed service-to-service communication.
Complete audit trail
Every significant action recorded with organization, team, session, service and environment. Risk assessment before sensitive actions.
Configurable policy engine
Tool-level permissions, rate limits, read-only mode, command allowlists, and automatic infinite-loop detection for runaway automation.
4-scope enterprise RAG
Persistent enterprise RAG memory: USER, PROJECT, ORG, and TENANT. Shared across agents and refreshed over time.
Automated quality gates
Pre-commit commands detect modified workspaces and run lint + tests. Quality gates remain inside the governed workflow.
Terminal, IDE, CI/CD, Docker
One binary, everywhere your team works.
Interactive TUI
Sidebar, agent tree, session history, and diff viewer in the terminal.
VSCode & JetBrains (ACP)
Native IDE extensions via Agent Client Protocol with files, diagnostics, LSP, and refactors from chat.
CI/CD — argy run
Headless autonomous mode for GitHub Actions, GitLab CI, and Azure Pipelines.
Docker
docker run -it --rm ghcr.io/demkada/argy-code — isolated execution without local install.
argy serve — HTTP API
Expose Argy Code as a local HTTP server for SDK integrations and custom automation.
Plugins & custom tools
.argy/plugin/*.ts, .argy/tools/*.ts, .argy/command/*.md, .argy/agent/*.md — project-level extensibility.
Core commands
The commands your team will use most often.
argy auth loginDevice Flow / PAT authenticationargy modelsModels available through LLM Gatewayargy mcp listConnected MCP serversargy agent listAvailable agentsargy lsp statusLanguage Server Protocol statusargy run "prompt"Autonomous executionargy session listSession historyargy serveLocal HTTP serverargy statsUsage and token statisticsArgy Code vs. alternatives
Why Argy Code is fundamentally different.
| Capability | Argy Code | GitHub Copilot | Cursor | Cline | Windsurf |
|---|---|---|---|---|---|
| Governed parallel multi-agent orchestration | ✓ native | ~ Agent Mode | ~ partial | ~ partial | ~ partial |
| Multi-day project continuity | ✓ | ✗ | ✗ | ✗ | ✗ |
| Default-enabled LSP | ✓ native | IDE only | IDE only | ✗ | IDE only |
| Automated security audit (8 phases) | ✓ | ✗ | ✗ | ✗ | ✗ |
| Browser automation (E2E, WCAG, Electron) | ✓ | ✗ | ✗ | ✗ | ✗ |
| Document generation (.pptx, .xlsx, .pdf, .docx) | ✓ | ✗ | ✗ | ✗ | ✗ |
| Enterprise multi-scope RAG (USER/ORG/TENANT) | ✓ | ✗ | ✗ | ✗ | ✗ |
| ACP (native Argy IDE integration) | ✓ | ✗ | ✗ | ✗ | ✗ |
| MCP (external tools) | ✓ | ✓ | ✓ | ✓ | ✓ |
| Reusable team skills | ✓ | ✗ | ✗ | ✗ | ✗ |
| Enterprise governance (doctrine, audit, multi-tenant) | ✓ native | ~ partial | ✗ | ✗ | ✗ |
| Autonomous CI/CD headless execution | ✓ | ~ Agent Mode | ~ CLI beta | ~ partial | ~ partial |
| Token optimization & checkpoint/resume | ✓ | ✗ | ✗ | ✗ | ✗ |
| Multi-provider LLM-agnostic | ✓ | ~ multi-model | ~ multi-model | ✓ | ~ multi-model |
FAQs
Common questions about Argy Code.
What exactly is Argy Code?⌃
Argy Code is an enterprise AI development agent that orchestrates a team of extensible specialist agents in parallel (Orchestrator, Builder, Planner, Explorer, Code-Reviewer, Sec-Auditor, Doc-Generator, Browser, Deep-Researcher, Skill-Creator, General) with 30+ built-in tools. It supports interactive TUI sessions and autonomous argy run execution for CI/CD.
How does Argy Code sustain complex multi-day work?⌃
Argy Code plans, delegates, validates and resumes work across sessions — with persistent checkpoints, progress summaries and human approvals before risky decisions. Long-running initiatives continue across several days without losing context.
How is governance enforced?⌃
Through ARGY.md project doctrine, the LLM Gateway (quotas, PII/secrets/prompt-injection filters, audit), no-privilege-escalation permission inheritance, tenant isolation, rate limits, and risk assessment before sensitive actions for sensitive actions.
What is LSP in Argy Code?⌃
Language Server Protocol (LSP) is enabled by default in Argy Code. It provides real-time code intelligence: diagnostics, symbols, reference search, go-to-definition, refactors, and semantic project understanding.
ACP and MCP: what is the difference?⌃
ACP (Agent Client Protocol) powers Argy Code IDE integrations such as VSCode and JetBrains. MCP (Model Context Protocol) connects Argy Code to standardized external tool servers — databases, APIs, DevOps systems, and internal services.
How does enterprise RAG memory work?⌃
Argy Code uses persistent vector memory across four scopes: USER, PROJECT, ORG, and TENANT. Agents query that memory before acting to avoid repeated mistakes and follow established codebase patterns.
How does token optimization work?⌃
Argy Code automatically compacts context at 80% of the window, preserves critical session context, and applies a multi-layer governance optimization. Checkpoint/resume recovers partial work after interruptions.
Install
Pick your platform — all methods install the same binary.
European SaaS
GDPR compliant & hosted in EU
No Lock-in
Built on open standards
API-First
Everything is automatable
Ready to get started with Argy?
Start with the Free plan. Upgrade when you're ready, or contact us for an enterprise rollout.