skawld

skawld / skawld-agent

Public

Skawld Agent Core & SDK

18
0
89% credibility
Found May 27, 2026 at 18 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

skawld is an open-source TypeScript framework that lets developers embed a full AI agent into their applications. The agent can use tools to read files, write code, run commands, and handle complex multi-step tasks while maintaining conversation history and respecting permission boundaries.

How It Works

1
💡 Discover skawld

You hear about skawld as a simple way to add an AI assistant to your application that can read files, run commands, and help with coding tasks.

2
📦 Install the package

You add skawld to your project with a single command, just like adding any other library.

3
🔑 Connect your AI service

You tell skawld which AI provider to use by setting your API key, then your assistant can think and reason just like a developer.

4
🤖 Create your agent

You set up your assistant with the tools it needs — like the ability to read and write files, run terminal commands, or search through code.

5
💬 Start a conversation

You open a session and ask your assistant to help with a task, like exploring a codebase or making changes.

6
🔧 Watch it work

Your assistant thinks, decides which tools to use, and works through your task step by step — reading files, making edits, running tests.

✅ Get your results

Your task is complete! The agent finished the work and you can see everything it did along the way.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 18 to 18 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is skawld-agent?

Skawld is a TypeScript agent framework that runs exclusively on Bun. It gives you a single import to embed a full LLM-powered agent loop into any application -- tools, streaming events, session persistence, and permission controls all bundled together. You create an Agent, attach a model provider (Anthropic or OpenAI), register tools, and iterate over the event stream. The framework handles the messy parts: tool scheduling, parallel execution, context compaction when conversations get long, and automatic retry on rate limits or context errors.

Why is it gaining traction?

The hook is simplicity. Compared to building this yourself or stitching together multiple libraries, skawld gives you a coherent surface in about twenty lines of code. The event-driven design means you get full visibility into what the agent is doing -- partial responses stream token by token, tool calls emit start/end events, and permission requests bubble up for user decisions. The built-in permission engine is also notable: write operations require explicit approval by default, while read operations auto-allow. That boundary is configurable but sensible out of the box.

Who should use this?

Backend developers building Bun applications who need LLM agents with file system access, shell execution, or integration with external tools via MCP. Teams evaluating agent frameworks for code generation, automated testing, or data processing pipelines will find the SDK surface approachable. If you need MCP server integration or subagent parallelism, skawld has those features natively. Early adopters comfortable with pre-1.0 software will get the most value.

Verdict

Skawld is a well-structured agent SDK with thoughtful defaults, but the credibility score of 0.9% and 18 stars tell you everything about its maturity. The codebase is tested and the API is coherent, but the community is essentially nonexistent. Use it for side projects or internal tooling where you can absorb breakage. For production systems, wait for a more established ecosystem or contribute meaningfully to accelerate that.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.