Pangolin
Version 0.8.0. A Rust Iceberg REST catalog with multi-tenancy, Git-style branching, RBAC, OIDC, encrypted warehouse credentials, audit logging, a management UI, two CLIs, and a Python SDK.
- Apache Iceberg
- Agentic AI
- Open Source
Alex Merced
Open Lakehouse & AI Advocate · Author · Technologist
I build and teach the tools behind the open data lakehouse and agentic AI, from Apache Iceberg to local-first AI agents.
As published by O'Reilly · Manning · Packt · LinkedIn Learning
01 / Watch
Three stories. One open future.
Meet Alex, open the agentic stack, and travel through the Apache Iceberg lakehouse in three original, scene-rich music videos.
02 / About
About Me
Alex Merced is a technologist, author, and one of the most active advocates for the open data lakehouse and agentic AI. He writes, builds, and teaches across the modern data stack: Apache Iceberg, Apache Polaris, query engines, and the AI agents that sit on top of them.
He is the author of four books with O'Reilly, Manning, and Packt, a LinkedIn Learning instructor, creator of DataLakehouseHub.com, and host of the Datanation and Web Dev 101 podcasts. In his day job he serves as Head of Developer Relations at Dremio, where he was the inaugural winner of the 2025 CEO Award. Alongside the writing and speaking, he ships open-source software, most recently the Mag ecosystem of local-first AI agent tooling.
- 34K+LinkedIn Followers
- 4Books · O'Reilly, Manning & Packt
- 575+Public GitHub Repos
- 300+GitHub Followers
- 10+Open-Source Projects
03 / Expertise
Technical Expertise
01
Data & Infrastructure
- Apache Iceberg
- Apache Polaris
- Data Lakehouse Architecture
- Dremio
- Docker & DataOps
02
AI & Agents
- Agentic AI & tool-use
- Model Context Protocol (MCP)
- RAG & vector search
- LLM app tooling (LiteLLM)
- AI agent memory / knowledge graphs
03
Languages
- JavaScript / TypeScript
- Python
- Rust
- Go
- SQL
04
Web & Frameworks
- React, Vue, Svelte, Angular
- Node.js (Express, Fastify)
- Python (FastAPI, Flask, Django)
04 / Open Source
Things I've Built
Open Lakehouse & Data Tooling
4 projectsOpen-source tools for Apache Iceberg catalogs, query engines, and the lakehouse workflows I write and teach about.
Apache Iceberg UI
A local desktop application for browsing and querying Iceberg Catalogs.
Iceframe
0.13.0 is published; 0.14.0 is in development. A DataFrame-style Apache Iceberg library with local Polars execution, table maintenance, broad ingestion, quality gates, async operations, and a bounded MCP surface. The development line adds MCP 2.1, complete typing, query explain, CDC, and correctness hardening.
Dremio Python CLI
A Python CLI for Dremio Software and Cloud.
Also built DremioFrame (dataframe library for Dremio Cloud) and SencilloDB 0.8.0 (a dependency-free embedded JSON document store with typed schemas, hardened transactions, atomic persistence, diagnostics, and selective shard writes).
Current status Pangolin, IceFrame, and SencilloDB each have a version, a maturity level, and a set of things they are not ready for. Read where each one stands
AI & Agents: the Mag Ecosystem
4 projectsA local-first AI productivity stack I build in the open: a Rust memory graph, a terminal agent, a desktop command center, and a broker for the agent CLIs you already run. It sits inside a larger effort to make agentic work portable and governable, alongside AGS, OAP, AAIS, and Loro.
MagGraph
Rust-powered graph database that gives AI agents a human-readable, Git-versioned memory.
An in-process graph database for AI semantic layers. Knowledge is stored as versioned Markdown nodes in a Git repo; edges emerge automatically from [[wikilinks]]. It ships a Python API, an auto-generated MCP server, and a CLI, plus a "lakehouse mode" where nodes point at external S3/Parquet data, so agents get compact recall bundles, backlinks, and traversal with zero infrastructure.
MagAgent
The memory-first personal agent: it remembers you across sessions, in Git-backed Markdown you can review.
Version 1.4.0. A CLI-first coding and productivity agent with MagGraph memory, 20+ model providers, reusable skills, subagents, gateways, OAP profiles, and AGS level 3 graph execution. Version 1.4 makes the memory inspectable: every turn records which memories were recalled, why they matched, and what they cost against the token budget, shown by /why last and a Memory used panel in the web UI. It also adds review-gated team memory, approval grants that expire and can be revoked, and an offline mock provider for first-run demos (experimental).
Mag Command Center
The desktop cockpit for MagAgent: runs, approvals, graphs, and memory in one window.
Version 1.0.0, the first stable release. A Tauri + React + TypeScript desktop app for Linux, macOS, and Windows that requires MagAgent 1.4.0. A new user can add a provider key or start an offline demo without a terminal, Stop ends every process a run started, a tray icon and OS notifications report waiting approvals, and a Memory used panel shows what MagAgent recalled for each run. The installers are unsigned, so macOS and Windows warn on first launch; each release publishes checksums to verify against.
Merced AI
One portable agent identity across the harnesses you already use, with honest reports of what each one drops.
Version 0.8.0. Deliberately not another agent loop. Merced AI discovers installed agent CLIs, normalizes their noninteractive interfaces, and binds OAP profiles to them as portable bots, while the selected harness stays the final authority. Version 0.8 runs Claude Code, Gemini CLI, Goose, and OpenCode over the Agent Client Protocol, gives each write-capable bot in a group its own Git worktree, compares harnesses on one profile and prompt, and lists every profile section a projection drops. An A2A endpoint is included as experimental.
The bigger picture MagAgent 1.4.0, Loro 0.22.0, Merced AI 0.8.0, and Mag Command Center 1.0.0 compose around three open contracts: AGS for work, OAP for agents, and AAIS for live human authorization. All three publish support libraries for Python, TypeScript, Go, Rust, and Java, and Loro, MagAgent, and Merced AI now compute OAP profile digests the same way. Read how they fit together
05 / Publications
Featured Books
-
Apache Iceberg: The Definitive GuideO'Reilly
-
Apache Polaris: The Definitive GuideO'Reilly
-
Architecting an Iceberg LakehouseManning
-
Building the Software FactoryPackt
I also host the Datanation and Web Dev 101 podcasts. The full reading list lives at books.alexmerced.com.
06 / Credentials
Recognition & Credentials
-
Technical Author
Published author with O'Reilly, Manning, and Packt, including Apache Iceberg: The Definitive Guide.
-
Talks & Speaking
Conference speaker at Oredev, Data Council, Data Day Texas, Confluent Current, SQLBits, P99Conf, DataEngBytes, OSA Con, and Subsurface. Available for keynotes and workshops.
-
Educator & Creator
LinkedIn Learning Instructor and creator of DataLakehouseHub, teaching data lakehouse and Apache Iceberg to a global developer community.
-
Open Source
Maintainer of 10+ open-source projects across the lakehouse and agentic-AI stacks, from the Pangolin catalog to the Mag ecosystem.
07 / Questions
Frequently Asked Questions
What does Alex Merced do?
Alex Merced builds, writes, and teaches about the open data lakehouse and agentic AI. He authors books and courses, speaks at conferences, hosts the Datanation and Web Dev 101 podcasts, ships open-source tooling, and serves as Head of Developer Relations at Dremio.
What is Alex Merced known for?
He is best known as one of the most active advocates and educators for Apache Iceberg and the open data lakehouse, as co-author of Apache Iceberg: The Definitive Guide, creator of DataLakehouseHub.com, and a LinkedIn Learning instructor. More recently he is known for building the open-source Mag ecosystem of local-first AI agent tooling.
What books has Alex Merced written?
Four books with major publishers: Apache Iceberg: The Definitive Guide (O'Reilly), Apache Polaris: The Definitive Guide (O'Reilly), Architecting an Apache Iceberg Lakehouse (Manning), and Building the Software Factory (Packt). A full reading list lives at books.alexmerced.com.
What open-source projects has Alex Merced built?
On the data side: Pangolin, Apache Iceberg UI, Iceframe, DremioFrame, the Dremio Python CLI, and SencilloDB. On the AI side, the Mag ecosystem covers MagGraph, MagAgent 1.4.0, Mag Command Center 1.0.0, Loro 0.22.0, Merced AI 0.8.0, and the open AGS, OAP, and AAIS specifications.
What is the Mag ecosystem?
The Mag ecosystem is a local-first AI productivity stack Alex builds in the open: MagGraph stores agent memory as Git-versioned Markdown nodes in a Rust graph database, MagAgent is the terminal agent that uses that memory across 20+ LLM providers, and Mag Command Center is the cross-platform desktop cockpit for both. Merced AI sits alongside them as a broker that runs the same Open Agent Profiles on whichever harness is already installed.
Where can I book Alex Merced to speak?
Alex speaks at conferences worldwide, including Oredev, Data Council, Data Day Texas, Confluent Current, SQLBits, P99Conf, DataEngBytes, OSA Con, and Subsurface. For keynotes, workshops, and panels on the open lakehouse, Apache Iceberg, or agentic AI, email Contact@alexmerced.com.
08 / Contact
Get in Touch
Looking for a conference speaker, a collaborator, or just want to talk about Apache Iceberg and AI agents? Here's how to reach me.
Book Me to Speak
Keynotes, workshops, and panels on the open lakehouse, Apache Iceberg, and agentic AI.
Say Hello
Questions about the projects on this page, collaborations, or feedback on the books.
Read & Follow
Books, courses, newsletters, and everything else I publish across the web.