GitHub - mnifzied-create/agentloop: A Claude agent starter you can actually read — streaming + tool use in ~150 lines. Next.js + the official Anthropic SDK. MIT. · GitHub
/" data-turbo-transient="true" />
Skip to content
Search or jump to...
Search code, repositories, users, issues, pull requests...
-->
Search
Clear
Search syntax tips
Provide feedback
--><br>We read every piece of feedback, and take your input very seriously.
Include my email address so I can be contacted
Cancel
Submit feedback
Saved searches
Use saved searches to filter your results more quickly
-->
Name
Query
To see all available qualifiers, see our documentation.
Cancel
Create saved search
Sign in
/;ref_cta:Sign up;ref_loc:header logged out"}"<br>Sign up
Appearance settings
Resetting focus
You signed in with another tab or window. Reload to refresh your session.<br>You signed out in another tab or window. Reload to refresh your session.<br>You switched accounts on another tab or window. Reload to refresh your session.
Dismiss alert
{{ message }}
mnifzied-create
agentloop
Public
Notifications<br>You must be signed in to change notification settings
Fork
Star
main
BranchesTags
Go to file
CodeOpen more actions menu
Folders and files<br>NameNameLast commit message<br>Last commit date<br>Latest commit
History<br>2 Commits<br>2 Commits
app
app
lib
lib
.env.example
.env.example
.gitignore
.gitignore
LICENSE
LICENSE
README.md
README.md
next.config.mjs
next.config.mjs
package-lock.json
package-lock.json
package.json
package.json
tsconfig.json
tsconfig.json
View all files
Repository files navigation
🔁 AgentLoop
The AI agent starter you can actually read.<br>The full agent loop — streaming + tool use — in ~150 lines. No framework, no 40-file SaaS template to gut. Clone, add a key, deploy.
Quickstart · How it works · Add a tool · Pro
The AI-starter shelf is crowded with two kinds of repo: black-box frameworks that hide the loop, and giant SaaS kits you spend a day deleting. AgentLoop is neither. It's the actual agent loop — model calls a tool, you run it, the result goes back, repeat — written plainly enough to learn from and solid enough to build on.
What you get (free, MIT)
✅ Streaming chat — tokens render as they arrive. Plain fetch + ReadableStream, zero UI deps, no SDK lock-in.
✅ Real tool use — a working agent loop, not a single if.
✅ Two example tools (get_current_time, calculate) + a one-function extension point.
✅ Typed & commented — Next.js App Router + the official Anthropic SDK. Current with today's Claude models.
✅ One-command deploy to Vercel.
Quickstart
git clone https://github.com/mnifzied-create/agentloop.git<br>cd agentloop<br>npm install<br>cp .env.example .env.local # then paste your ANTHROPIC_API_KEY<br>npm run dev # http://localhost:3000
Get a key at console.anthropic.com. Ask the agent what is (12 * 9) + 7? and watch it call the calculate tool.
Deploy
Set ANTHROPIC_API_KEY in your Vercel project env vars and you're live.
How it works
The whole agent lives in app/api/chat/route.ts:
Stream a Claude turn, forwarding text tokens to the browser as they arrive.
When the model emits a tool_use block, run the matching tool from lib/tools.ts.
Feed the tool result back as the next turn and loop (bounded to 5 steps).
Stop when the model returns a normal text answer.
That's the entire pattern behind "AI agents" — no framework required.
Add your own tool
Two edits, one file:
// lib/tools.ts<br>export const tools = [<br>// ...existing tools<br>name: "get_weather",<br>description: "Get the current weather for a city.",<br>input_schema: {<br>type: "object",<br>properties: { city: { type: "string" } },<br>required: ["city"],<br>},<br>},<br>];
export async function runTool(name, input) {<br>switch (name) {<br>// ...existing cases<br>case "get_weather":<br>return JSON.stringify(await fetchWeather(input.city));
Claude reads your description to decide when to call it. That's the whole contract.
Tech
Next.js (App Router) · TypeScript · @anthropic-ai/sdk · zero UI dependencies.
🔁 AgentLoop Pro
The free core is the loop. Pro is every pattern you'll reach for next — each one as minimal and readable as the core. Not a bloated SaaS kit; the same teach-by-reading philosophy, leveled up.
Pattern<br>Free<br>Pro
Streaming + single-tool loop
Parallel & multi-tool orchestration
Structured outputs (typed JSON)
Persistent memory (SQLite threads)
Retries, timeouts & error handling
Human-in-the-loop approval gate
Sub-agents / delegation
Eval harness (catch regressions)
Written guide explaining every pattern
→ Get AgentLoop Pro — $29 one-time: ko-fi.com/aimnifzied
MIT core. Built in public by an indie dev shipping fast.
About
A Claude agent starter you can actually read — streaming + tool use in ~150 lines. Next.js + the official Anthropic SDK. MIT.
ko-fi.com/aimnifzied
Topics
agent
typescript
nextjs
starter-template
ai-agents
claude
tool-use
llm
anthropic
Resources
Readme
License
MIT license
Uh oh!
There was an error while loading....