SafeDusk

For people interested in rolling their own with TUI interface, I've an open sourced, single file implementation here[0].

It is based off Shopify CEO Tobi's `try` implementation[1].

[0]: https://github.com/aperoc/toolkami [1]: https://github.com/tobi/try

Resonating with the statement "Designing agentic loops is a very new skill—Claude Code was first released in just February 2025!" when I was investigating Codex[0].

[0]: https://blog.toolkami.com/openai-codex-tools/#coding-agents

This is super interesting! I've dabbled with Perlin noise procedural generation using AlphaEvolve[0] and wonder if it would be interesting to do one with Voronoi map too!

[0]: https://blog.toolkami.com/alphaevolve-toolkami-style/

Trying to figure out how I can make agentic development better with: https://toolkami.com

The space is moving so fast, so I had to note down my thoughts e.g. https://blog.toolkami.com/openai-codex-tools/and figure out what's next.

I'm actually curious about how you all see coding agents play out when every major model provider also provides a companion coding agent e.g. Claude Code, OpenAI Codex, Gemini cli ...

I wrote my thoughts[0] and felt whilst it is good to be model agnostic, the core has to be open because it is too important a tool to be left entirely in hand of others. It is also super simple to understand, OpenAI only has <5 tools for a coding agent to be functional.

[0]: https://blog.toolkami.com/openai-codex-tools/

Definitely keep it simple, and OpenAI Codex's tooling setup[0] agrees with you too! Have a simple `update_plan` to track items in progress.

[0]: https://blog.toolkami.com/openai-codex-tools/

OpenAI Codex has an `update_plan` function[0]. I'm wondering if switching the implementation to this would improve the coding agent's capabilities or is the default for simplicity better.

[0]: https://blog.toolkami.com/openai-codex-tools/

Terminal and shell are becoming more and more important with the advent of coding agents, the graphical progress bar is a super nice touch too

Agents enabled me to quickly implement and test out papers by treating them as LLM workflows with composable tools instead of direct implementation.

I have a <a href="https://blog.toolkami.com/alphaevolve-toolkami-style/">blog post</a> detailing this.

[1]: https://blog.toolkami.com/alphaevolve-toolkami-style/

Creating your own MCP server in Cursor requires <50 lines of code for the base (https://blog.toolkami.com/toolkami-shttp-server/)

I am using this feature in my MCP servers and it is a god send! Some details on how I do it can be found at https://blog.toolkami.com/toolkami-shttp-server/.

Hi HN, I've written a blog post on how to create a simple sHTTP MCP server and have Cursor connect to it.

The server is <50 lines of code and have a few interesting features: - You can effectively run this program like a binary with all dependencies through UV script - You can selectively decide which tool makes sense for your current application as they are mounted on different paths Mount("/filesystem/"), ...) - Your agent can self-edit its tool in realtime because uvicorn’s reload=True and reload_dirs=["."] settings hot-reload modules whenever there is a change

Hope you find it useful!

Pretty close to what I discovered is essential in https://github.com/aperoc/toolkami, 7 tools will cover majority of the use cases.

Glad to see them working on this, as a Gemini user, I had to build https://github.com/aperoc/toolkami for myself before this.

Glad to see they're joining the game, there is so much work to do here. Have been using Gemini 2.5 pro as an autonomous coding agent for a while because it is free. Their work with AlphaEvolve is also pushing the edge - I did a small write up on AlphaEvolve with agentic workflow here: https://toolkami.com/alphaevolve-toolkami-style/

Instead of connecting to a server with 1000(s) of tools, I'm going the opposite direction and claim that you only need <10 sharp tools/small function for most use cases.

As an example, today I re-implemented Google's AlphaEvolve with <7 tools (https://toolkami.com/alphaevolve-toolkami-style/).

Exciting to see a fellow builder in the space! Love how memory management and observability is built-in.

Today, I just re-implemented Google's AlphaEvolve for Perlin noises using my own minimal agentic framework (https://toolkami.com/alphaevolve-toolkami-style/), will probably steal a trick or two from you.

I implemented a simplified version of Google's AlphaEvolve to optimize Perlin noise code to resemble a target fire image.

More details at: https://toolkami.com/alphaevolve-toolkami-style/

Sharing an agent framework (from scratch) that works very well with just 7 composable tools: read, write, diff, browse, command, think and ask.

Just pushed an update this week for OpenAI-compatibility too!

https://github.com/aperoc/toolkami

I am betting on myself.

I built a minimal agentic framework (with editing capability) that works for a lot of my tasks with just seven tools: read, write, diff, browse, command, ask and think.

One thing I'm proud of is the ability to have it be more proactive in making changes and taking next action by just disabling the `ask` tool.

I won't say it is better than any of the VSCode forks, but it works for 70% of my tasks in an understandable manner. As for the remaining stuff, I can always use Cursor/Windsurf in a complementary manner.

It is open, have a look at https://github.com/aperoc/toolkami if it interests you.

Changing my old coding behavior aside, biggest limiting factor for me is understanding how and why the coding agent is doing this a certain way, so that I have the confidence to continually sharpen my tools.

I want something simple that I have full control on, if not just to understand how they work. So I made a minimal coding agent (with edit capability) that is fully functional using only seven tools: read, write, diff, browse, command, ask, and think.

As an example, I can just disable `ask` tool to have it easily go full autonomous on certain tasks. Or, ask it to `think` for refactoring.

Have a look at https://github.com/aperoc/toolkami to see if it might be useful for you.

I've used both Cursor and Aider but I've always wanted something simple that I have full control on, if not just to understand how they work. So I made a minimal coding agent (with edit capability) that is fully functional using only seven tools: read, write, diff, browse, command, ask, and think.

I can just disable `ask` tool for example to have it easily go full autonomous on certain tasks.

Have a look at https://github.com/aperoc/toolkami to see if it might be useful for you.

It is based off https://www.anthropic.com/engineering/claude-think-tool.

For sure! I'm surprised myself how far one can get with just seven tools: read, write, diff, browse, command, ask, and think.

It's like lisp's original seven operators: quote, atom, eq, car, cdr, cons and cond.

And I still can't stop smiling just watching the agent go full turbo mode when I disable the `ask` tool.

Ideally, just start the server - which is a SSE based and supported by any MCP client out of the box.

Then, connect it using this line: `client = MCPClient(server_url=server_url)` (https://github.com/aperoc/toolkami/blob/e49d3797e6122fb54ddd...)

Happy to help further if you run into issues.

That is the dream! Would love someone to create a vim plugin for this, if not I'll do it myself if there is enough demand.

Tried Zed and Cursor, but they always felt too magical to me. I ended up building a minimal agent framework that only uses seven tools (even for code edits): read, write, diff, browse, command, ask, and think.

These simple, composable tools can be utilized well enough by increasingly powerful LLM(s), especially Gemini 2.5 pro to achieve most tasks in a consistent, understandable way.

More importantly - I can just switch off the 'ask' tool for the agent to go full turbo mode without frequent manual confirmation.

I just released it yesterday, have a look at https://github.com/aperoc/toolkami for the implementation if you think it is useful for you!

For sure! the traditional craftsman in me still like to do some stuff manually though haha

Yes, uv just feels so magical that I can't stop using it. I want to create the same experience with this!

Hope you find it useful, feel free to reach out if you need help or think it can be made better.