AI agents that actually work
Spawn persistent agents with memory, skills, plugins, and controlled system access. They learn, collaborate, and get things done — on your hardware.
Everything agents need to be useful
A complete runtime with persistent memory, powerful tools, skill learning, plugins, and multi-agent orchestration.
Persistent Memory
Agents remember context across sessions. Knowledge is stored, indexed, and recalled automatically.
30+ Built-in Tools
Shell execution, code running, file management, web fetching, scheduling — all out of the box.
Self-Learning Skills
Agents discover reusable procedures and propose saving them as skills for future use.
Multi-Agent Collaboration
Spawn specialized agents that communicate, delegate, and work together on complex tasks.
Human-in-the-Loop
Every destructive action requires approval. You stay in control while agents stay productive.
Runs Locally
Runs entirely on your hardware. Your data stays on your machine — only LLM requests reach your chosen provider.
Scheduling & Automation
Set up recurring tasks with interval or crontab schedules. Background monitoring that runs autonomously.
LLM Agnostic
Works with Anthropic, OpenAI, Google, Mistral, xAI, DeepSeek, Groq, and any OpenAI-compatible provider. Switch models anytime.
Encrypted Secrets
Store API keys and credentials encrypted at rest. Scope secrets globally or per agent.
Webhook Integration
Trigger agents from external services via HTTP webhooks with HMAC-SHA256 signature verification.
Durable Execution
Agents survive restarts and restore conversation context. Persistent state means nothing is lost.
Workspace Brain
Share facts, procedures, and preferences across all agents through a central knowledge base.
Smart Model Routing
Classifies task complexity and routes simple requests to faster, cheaper models automatically.
Vision Support
Attach images and screenshots. Vision-capable models analyze visual content directly from your workspace.
Usage Tracking
Monitor token consumption, LLM calls, and cost per agent. Set step, time, and token budgets.
Context Compaction
Automatically summarizes long conversations to stay within context limits while preserving key information.
Plugin System
Extend agent behavior with lifecycle hooks and custom plugins. Auto-discovered from workspace, agent, or global directories.
Docker Sandbox
Run code in isolated Docker containers for safe execution. Full environment control with automatic cleanup.
Capability Gap Detection
Agents detect when they lack tools for a task and request new capabilities instead of giving up.
Dry-Run Mode
Preview agent plans without executing any tool calls. Validate approaches before committing to real actions.
Built-in tools for real work
Agents ship with a powerful set of tools — from shell access to inter-agent messaging.
…and 20+ more including workspace search, heartbeat monitoring, structured prompts, and reasoning tools.
26 agent templates, ready to go
Pre-configured agents for common roles. Customize the soul, bootstrap, and skills to fit your workflow.
Templates are automation starting points, not professional services. They do not constitute legal, financial, tax, or medical advice.
Modular by design
Four layers that compose into a self-contained agent platform.
Web UI & Desktop App
Next.js dashboard and Electron app. Chat with agents, review approvals, manage workspace files.
Gateway
Central API server. Routes messages, manages sessions, supervises agent workers, serves the web UI.
Agent Runtime
The core. Runs the agent loop with plugin hooks, manages context, calls LLMs, executes tools, and compacts history.
Storage & Brain
SQLite-based persistence. Event sourcing, embeddings, vector search, long-term memory retrieval.
Download the Desktop App
Get WerkLane running in seconds. Configure your LLM provider and start spawning agents.