Reported News
NamiAgent: from-scratch local AI agent tutorial with MCP and memory
NamiAgent is both a zero-basics AI agent tutorial and a local runnable agent. The README says it starts from a 100-line loop, adds tools, permissions, MCP, sub-agents, planning and memory across lessons 01–11 (49 tests by lesson 11 with a fake model), and opens a local web UI. When we checked on Nov 7, 2026 the GitHub repo had 400 stars and 41 forks.
- Changes reported
- 6
- listed below, each from a source we read
- We tested
- Not tested yet
- reported from sources; we did not run it ourselves
- Gaps we flag
- 3
- things we could not verify
- Sources
- 2
- 1 primary, linked below
What changed
| # | What the source says | Status |
|---|---|---|
| 01 | New open-source project: aa1143/NamiAgent was created on GitHub on October 30, 2026 as a beginner AI-agent tutorial that is also a local runnable agent. The README describes file read/write, command execution, MCP, long-term memory and a built-in web UI that shows each step. | Reported |
| 02 | Lesson path: lessons/lesson-01 through lesson-11 are runnable snapshots. Parts 1–2 cover the loop, tools, permissions, context, Skills, chat apps, scheduler, MCP, sub-agents, planning and memory; part 3 (articles only) dissects Claude Code, Codex, Pi, Grok Bot, DeepSeek Harness, OpenClaw and Hermes. | Reported |
| 03 | Concrete scope from the README: starts from a 100-line minimal loop; lesson 11 has 49 tests run with a fake model (no API key); depends only on the openai package with no agent framework; Python 3.10+; OpenAI-compatible APIs, defaulting to DeepSeek. | Reported |
| 04 | Local web UI: python web/app.py (or --demo without an API key) serves http://localhost:8765, showing model turns and tool calls on the left and the full context on the right, with permission popups for file writes and commands. | Reported |
| 05 | Safety and positioning: from lesson 3 the agent is limited to NAMI_WORKSPACE and asks before side effects. The README states it is a readable minimal agent, not a production framework, and is not meant to match Claude Code or Codex feature-for-feature. Code is MIT; docs/ articles are CC BY-NC 4.0. | Reported |
| 06 | Traction when checked: the GitHub API reported 400 stars and 41 forks on Nov 7, 2026 (~8 days after creation). Companion docs and visualization live at namiai.tech/agent/ and namiai.tech/matrix/agent/. | Reported |
reported by the sourcemeasured or run by usnot verified
Primary source: NamiAgent repository and README (GitHub). We are the fast, independent index; the vendor’s own page is the source of record for this release.
In the source’s words
“不只是教程，也是一个能在本地跑起来的 Agent，自带网页端，每一步都看得见”
“从 100 行的最小循环开始，一课加一个能力：工具、权限、上下文管理、Skills、MCP、子 Agent、计划、记忆。”
“第 3 课起每课都有测试，第 11 课共 49 个，用假模型跑，不需要 API Key”
Early verdict
Worth a look if you want a Chinese/English-friendly from-scratch agent curriculum with runnable lesson snapshots and a local web UI; skip it if you need a production coding-agent product — the README positions it as teaching code.
Basis: Based only on the GitHub README and namiai.tech/agent/ as read on Nov 7, 2026.
What we have not verified
- We did not clone, install or run NamiAgent, open the web UI, or execute the lesson-11 test suite.
- We did not verify DeepSeek or other OpenAI-compatible API behaviour, Telegram bot setup, or MCP server integration beyond the README claims.
- We did not read every dissection article (lessons 12–18) or confirm those product details against the vendors' current docs.
This brief is Reported: written from the sources below, not from running the release ourselves.
Sources
- NamiAgent repository and README (GitHub)primary source · fetched 2026-11-07
- From-scratch Agent tutorial series (namiai.tech)source · fetched 2026-11-07
Related
Spotted an error? See our corrections policy or email the address on the contact page. Markdown version for agents: /news/namiagent-from-scratch-ai-agent-tutorial.md
Want these briefs in your Google Top Stories? Add Agentic Weekly as a preferred source on Google.
Newsletter
The Weekly Verdict. Free, every Monday.
A free email every Monday: the week's few big AI releases, then 5 short use cases for the new models, each with a prompt you can paste.
Free. Double opt-in. One click to unsubscribe.