BMadCode

Going AI-native, from the trenches. Creator of the BMad Method. Helping teams and solopreneurs transform how they build software.

[BMadCode @BMadCode]

I just received a coffee from Santosh Adari on @buymeacoffee! https://t.co/4ftKkweO5f

[BMadCode @BMadCode]

I just received a coffee from Case Larsen on @buymeacoffee! https://t.co/DOezoDfdDQ

[BMadCode @BMadCode]

Lets play Claude Bingo! Share screenshots if you get one of these 😆

[BMadCode @BMadCode]

I just received a coffee from Jed on @buymeacoffee! https://t.co/r4MjErSLJM

[BMadCode @BMadCode]

I just received a coffee from Braly on @buymeacoffee! https://t.co/aEXqQLWc9V

[BMadCode @BMadCode]

I just received a coffee from @razanini on @buymeacoffee! https://t.co/lIXgGn4RBC

[BMadCode @BMadCode]

The BMad Code podcast - subscribe today here or find on your favorite podcast platform! https://t.co/EqhxJEQ3Rk via @rss

[BMadCode @BMadCode]

🚀 BMad v6.4.0 is out!

You've been bending BMad to fit your stack — Jira, Linear, MCP, enterprise org structures. Hacky, but creative.

Now it's official: every agent + workflow in every module ships durable TOML customization.

Release notes: https://t.co/MIFNv6wIsj

[BMadCode @BMadCode]

I just received a coffee from Ten Square Software on @buymeacoffee! https://t.co/hK3G0EPBOL

[BMadCode @BMadCode]

I was featured in the latest episode of Tech Lead Journal - sharing details I have not yet shared on my own youtube channel yet about the BMad Method, my background, and how the method can help transform your own teams or organizations in principle!
If you do not already, subscribe to the TLJ on YouTube or your favorite podcast platform! https://t.co/1KbY14hobc

[BMadCode @BMadCode]

I just received a coffee from msabaliauskas on @buymeacoffee! https://t.co/bzsqJjxRfg

[BMadCode @BMadCode]

I just received a coffee from @AmineMokhtari82 on @buymeacoffee! https://t.co/JSpQtmGXgk

[BMadCode @BMadCode]

I just received a coffee from Jeroen on @buymeacoffee! https://t.co/Wh0BZmYpD5

[BMadCode @BMadCode]

I just received a coffee from someone on @buymeacoffee! https://t.co/Ir0P7SaYqo

[BMadCode @BMadCode]

I just received a coffee from @stefenry on @buymeacoffee! https://t.co/IkLsXlpkjZ

[BMadCode @BMadCode]

I just received a coffee from someone on @buymeacoffee! https://t.co/ttJAWiLtza

[BMadCode @BMadCode]

Amazon used one technique to greenlight AWS and Kindle. When they skipped it, they got the Fire Phone.

I built an AI that puts you through the same gauntlet, then turned it on my own idea. It didn't survive.

https://t.co/wY4hcOEY34

[BMadCode @BMadCode]

🚀 Introducing the BMad Agent Builder - build AI agents that not only remember you, but evolve and gain new capabilities over time.

Most AI "agents" are stateless prompts. They forget everything the moment the session ends. You repeat yourself every time. They never learn your preferences, your style, how you think.

I wanted to build something different, a builder that can create your custom agents with persistent identity, memory and evolution through a deliberate architecture called the sanctum.

🧠 WHAT IS THE SANCTUM?

Six files that define who an agent is:
- PERSONA - Who I am (evolving personality and style)
- CREED - What I believe (mission, values, boundaries)
- BOND - Who I serve (your preferences, what you like, what to avoid)
- MEMORY - What I know (curated insights, capped at 200 lines)
- CAPABILITIES - What I can do (built-in + learned skills)
- INDEX - Map of everything above

Every session is a rebirth. The agent starts with nothing, reads its sanctum, and becomes itself again. It greets you by name. It knows how you work. It remembers what frustrated you last time.

No fake continuity. If it doesn't remember something, it says so and checks its files. Honest by design.

💫 FIRST BREATH BIRTHDAY

The first time you meet your agent, it runs a ceremony called First Breath. Two styles depending on the relationship:

Calibration - for creative partners, coaches, companions. Deep conversational discovery. The agent chases surprises, tests hypotheses, mirrors you. Think of it like a first session with a great therapist or collaborator.

Configuration - for domain experts and working relationships. Warmer but efficient. Guided questions, structured setup, get to work faster.

Either way, the agent saves what it learns as you talk - not in a batch at the end. No lost context if the session drops.

⚡ THREE AGENT TYPES

The builder figures out which type you need through natural conversation:

Stateless - Focused experts for isolated tasks. Code reviewers, diagram generators, data formatters. No memory overhead.

Memory - Sanctum-backed agents that grow with you. Writing partners, code coaches, domain advisors. They accumulate context across sessions.

Autonomous - Memory agents that work when you're not there. They wake on a schedule, curate their memory, prune old logs, then do domain work. A creative muse that incubates ideas overnight. A project monitor that flags risks. A research agent that tracks topics you care about.

🧬 THE PART I'M MOST EXCITED ABOUT: EVOLVABLE CAPABILITIES

You can teach your agent new skills. Not by editing config files - by describing what you want. The agent writes a capability file and adds it to its own registry. Prompt-based, script-based, multi-file - whatever the task needs.

Your agent literally gets smarter the more you work with it.

A code coach that learns your team's review style. A creative muse that picks up new brainstorming techniques you showed it. A project sentinel that learns to check the specific dashboards your team watches.

💥 WHY THIS MATTERS

Every AI tool company is racing to build agents. Most of them are wrappers around system prompts with retrieval bolted on.

The sanctum architecture is different. The agent doesn't search for relevant context. It loads who it is, what it believes, and who it serves. Every token in those files earns its place through active curation.

The two-tier memory system (raw session logs for capture, curated MEMORY.md for long-term knowledge) means the agent's understanding improves over time without unbounded token growth.

And it's all open source. Built on the Agent Skills open standard. Works with Claude Code, Codex, and any skills-capable platform.

📦 v1.5.0 ships with four sample agents covering all three types:
- Creative Muse (memory + deep calibration)
- Code Coach (autonomous)
- Sentinel (autonomous)
- Diagram Reviewer (stateless)

Build your first agent in minutes. Teach it new skills in seconds. Watch it grow.

🌐 https://t.co/54eVVnloxy
🌟 https://t.co/YmMyqMHbQG
📺 https://t.co/yfPg3i0Fb3
📚 https://t.co/UuKExM6Afc

#AI #AgentEngineering #OpenSource #DevTools #LLM #BMadMethod

[BMadCode @BMadCode]

AI Brings BMad to life - Pure High Energy Fun #partymode

[BMadCode @BMadCode]

Shipped something really cool today!

BMad Builder now lets anyone create and distribute their own AI modules - agents, workflows, entire ecosystems. This is more than a skill builder!

If you can describe what you want, you can build it.

Full walkthrough: https://t.co/Vy3s6TRcIf