Guides & Setup
| Document | Description |
|---|---|
| Getting Started | Install, quick start, configuration, CLI reference, task API |
| The Bernstein Way | Architecture tenets and default workflow philosophy |
| Linux Install Guide | Platform-specific installation notes for Linux |
| GitHub App Setup | Step-by-step GitHub App installation and webhook configuration |
| GitHub Action | CI-triggered orchestration via GitHub Actions |
Architecture & Design
| Document | Description |
|---|---|
| Concepts | Task server, spawner, janitor, agent lifecycle, model routing |
| Design Document | Full system architecture, component interactions, data flow |
| Dependency Resolver | How task dependency graphs are resolved and scheduled |
| RAG Challenge Swarm | Origin story: 12 agents, 47 hours, 737 tickets, architectural lessons |
Deployment
| Document | Description |
|---|---|
| Deployment Guide | Production deployment patterns, storage backends, environment config |
| Cluster Mode | Distributed multi-node orchestration with worker daemons |
| Helm Deployment | Kubernetes deployment via Helm charts |
Protocols & Compatibility
| Document | Description |
|---|---|
| Protocol Compatibility Matrix | MCP 1.0/1.1, A2A 0.2/0.3, ACP — test results and version support |
| Zero Lock-In | How Bernstein avoids provider and protocol lock-in |
Governance & Policy
| Document | Description |
|---|---|
| Model Routing Policy | Provider allow/deny lists, model selection rules, cost controls |
| Known Limitations | Current constraints, workarounds, and planned fixes |
Extensibility
| Document | Description |
|---|---|
| Adapters | Agent adapter reference: all supported CLIs, feature matrix, custom adapters |
| Plugin SDK | Pluggy-based plugin system: hooks, lifecycle events, packaging |
| API Reference | Task server HTTP API: endpoints, schemas, config |
Comparisons
| Document | Description |
|---|---|
| Comparison Index | Overview of all comparison pages with selection guidance |
| vs. Conductor | Workflow engine comparison |
| vs. Crystal | Multi-agent framework comparison |
| vs. Dorothy | Orchestration system comparison |
| vs. GitHub Agent HQ | GitHub's multi-agent system comparison |
| vs. Parallel Code | Parallel coding tool comparison |
| vs. Single Agent | Multi-agent orchestration vs. single-agent baseline |
| vs. Stoneforge | Agent platform comparison |
| Competitive Matrix | Feature-by-feature comparison table across tools |
Coverage & Status
| Document | Description |
|---|---|
| Feature Matrix | Documentation coverage status for every shipped capability |
| Changelog | Release notes and version history |
Design Decisions
| Document | Description |
|---|---|
| 001 — Agent Lifecycle | Why short-lived agents over long-running processes |
| 002 — UX and Distribution | CLI-first design, distribution strategy |
| 003 — Self-Evolution | Risk-gated self-improvement architecture |
Blog & Write-ups
| Document | Description |
|---|---|
| Zero-LLM Coordination | How Bernstein achieves multi-agent orchestration without LLM scheduling |
| Multi-Agent Benchmark | Benchmark results: orchestrated vs. single-agent performance |
| Self-Evolution: 30 Days | Results from 30 days of continuous self-improvement |
| SWE-Bench Orchestration | Applying orchestration patterns to SWE-bench tasks |