Technical deep dives and project updates from the team behind OpenCaddis and the FabrCore framework.
What we learned about LLM performance on Azure OpenAI while building an adaptive confidence-threshold reasoning agent that breaks complex questions into steps.
February 21, 2026 at 2:35 PM • 12 min read
What we learned about LLM performance while building an adaptive reasoning agent that breaks questions into steps and iterates until confident.
February 21, 2026 at 7:35 AM • 5 min read
Connect your agents to external tools like GitHub, Slack, databases, and file systems through MCP — no code changes required.
February 20, 2026 at 2:10 PM • 3 min read
The framework that powers OpenCaddis has been renamed. Here's the story behind the change.
February 20, 2026 at 11:41 AM • 6 min read
Add custom named model configs from the Settings UI and assign different providers to different agents.
February 20, 2026 at 9:55 AM • 7 min read
Reminders now carry explicit perspective — notify the user or act autonomously — with one-shot defaults and reliable UI delivery.
February 20, 2026 at 8:40 AM • 8 min read
The initial release of OpenCaddis.Sdk — auto-discovery, persistent configuration, and the first step toward an addon marketplace.
February 19, 2026 at 9:31 AM • 12 min read
A deep dive into automatic conversation summarization — the algorithm, safety mechanisms, and configuration options behind compaction.
February 18, 2026 at 11:03 AM • 14 min read
How we built an LLM-powered router that selects the best agent for each request, formulates optimized messages, and handles timeouts gracefully.
February 17, 2026 at 9:47 AM • 16 min read
From its origins as a derivative of the OpenClaw Lobster plugin to a full pipeline engine — DSL parsing, safety validation, approval gates, and parallel execution.
February 16, 2026 at 10:22 AM • 8 min read
Announcing the open source release of OpenCaddis and FabrCore — why now, what's included, and a note of gratitude to the community that made it possible.