Spacebot (spacebot.sh) just landed on Hacker News as a direct alternative to OpenClaw — but with a fundamentally different architecture. Instead of treating AI agents as monolithic chat interfaces, Spacebot splits the LLM process into three dedicated roles: Channel (the user-facing ambassador with personality), Branch (a context fork for thinking), and Worker (focused execution without conversation overhead). That's infrastructure-level thinking, not another wrapper around GPT.

Role-Based Architecture

The Channel is what users actually talk to — one per conversation, with full identity and personality. It delegates everything else. When the Channel needs to think through something complex, it forks a Branch with the full conversation history; the Branch chews on it and returns only the conclusion. Meanwhile, Workers do the actual work — they get a task, the right tools, and zero personality. This isn't sequential processing; it's concurrent orchestration where nothing blocks.

The Cortex Memory System

Spacebot's Cortex is the real differentiator. Every 60 minutes, it queries the memory graph across 8 dimensions and synthesizes a briefing that every conversation inherits — lock-free, zero-copy. The Association Loop continuously scans memories for embedding similarity and builds graph edges between related knowledge. Facts link to decisions. Events link to goals. The system gets smarter automatically.

File Ingestion and OpenClaw Migration

Drop text files into the ingest folder — notes, docs, logs, markdown, whatever — and Spacebot chunks them through an LLM with memory tools to produce typed, graph-connected memories. For OpenClaw users migrating over, just drop your MEMORY.md and daily logs into the ingest folder; Spacebot extracts structured memories and wires them into the graph automatically. Supported formats: .md, .txt, .json, .yaml, .csv, .log, .toml, .xml, .html, .rst.

Tech Stack and Deployment

Built in Rust with Tokio, SQLite (redb), LanceDB, FastEmbed, Serenity, and Chromiumoxide. First-class support for 10 LLM providers with automatic routing and fallbacks. Single binary, no runtime dependencies, deployable via Docker on any platform — AWS, GCP, bare metal, private networks. Self-host with one command: docker run -d --name spacebot -v spacebot-data:/data -p 19898:19898 ghcr.io/spacedriveapp/spacebot:latest

Pricing and Support Plans

Hosted plans: $29/mo (3 agents), $59/mo (6 agents, most popular), $129/mo (12 agents). Self-hosted support: $59/mo basic or $299/mo priority. Enterprise contracts with SLAs available. All plans require your own LLM API keys (BYOK) — bundled credits coming soon.

Key Takeaways

  • Three dedicated LLM roles (Channel/Branch/Worker) enable concurrent non-blocking AI processes
  • Cortex memory synthesis provides pre-computed briefings so conversations never start cold
  • Association Loop builds graph edges automatically — memory gets smarter without manual tagging
  • Rust-built infrastructure means single binary deployment, no Docker beyond the container, no microservices sprawl
  • OpenClaw migration path is drop-in compatible — MEMORY.md and daily logs just work

The Bottom Line

Spacebot isn't a chatbot. It's infrastructure for teams that need autonomous AI processes with shared mutable state, persistent memory across conversations, and zero human oversight. The role-based LLM architecture is the right call — treating AI as a workforce rather than a fancy autocomplete. If you're building AIops or need real agentic workflows, this is worth a serious look. OpenClaw has competition now.