Today we are introducing two new features: Re-plan and Continuous Characterization. Re-plan allows the Task Agent to read, assess and edit plan documents made by Claude Code. It is a game-changing capability for the Task Agent, enabling a cavalcade of new workflows.
\nCharacterization is part of the secret sauce that makes Task Demon do what it does so well. And today, Characterization becomes Continuous Characterization, automatically updating your Characterization documents as you implement tasks. This ensures that as you add new code to your codebase, your Characterization documents are always up to date, so Task Demon's Agents can continue to make the best possible decisions.
\nCheck out the short video below to see both new features in action:
\nThe Task Agent got way smarter this week, gaining the ability to re-plan a task if it feels that the first plan wasn't good enough, or if you tell it that you want to change things. The Task Agent automatically reads plan documents the moment they are submitted, analyzes them and gives you a quick summary of what the plan proposes. Task Agent gives the plan a rating and provides detailed feedback on what needs to be improved.
\nIf it feels that a re-plan is warranted, it will do so automatically up to 3 times, though you can ask it to re-plan any time you like.
\nHere's an example of an actual task agent session where we re-plan a task. In this case I wanted a button that would regenerate the plan background document if it exists already, but I didn't like the first iteration of the plan so I asked it for some tweaks.
\nThe UI below is a read-only view of the conversation between the Task Agent and the human driving it (me). The buttons are not interactive, but you can scroll through the short conversation to see re-planning in action:
\nRe-plan is available in the Task Agent today for all users. Shipping as part of this update is the ability for Task Agent to read each version of the plan and answer your questions about it - try it out!
\nOne of the ways Task Demon stands apart is through its Characterization process - an automated multi-agentic process that automatically analyzes and characterizes your codebase. It's always recommended to run Characterization when you start a new project in Task Demon so that it knows how your project works, but what about changes you make after that?
\nThat's where Continuous Characterization comes in. Now, every time you click the \"Mark Implemented\" button on a Task, the Characterization Agent will be informed. It will be passed the latest plan document for the Task and will automatically decided whether any other existing Characterization documents need to be updated, new ones created, or for nothing to happen at all.
\nContinuous Characterization keeps your Characterization up to date automatically, without any effort on your part. It's critical for making sure that your Task Demon Characterization stays in sync with your actual code, and it's shipping today for all users.
\nThese two new features bring significant new versatility to the Task Demon way of building software, and also aid us in shipping more features faster. We'll be announcing a bunch of stuff over the coming weeks, so keep your eyes peeled, sign up for our newsletter if you haven't already, and head over to our Discord channel to hang out and ask us anything.
\nIn the meantime, we have free trials available and the individual plan is only $15/month, so what are you waiting for?
", "url": "https://taskdemon.ai//blog/replan-and-continuous-characterization", "title": "Introducing Re-plan and Continuous Characterization", "image": "https://taskdemon.ai/images/posts/2025/replan-and-continuous-characterization.png", "date_modified": "2025-04-27T19:27:09.000Z" }, { "id": "code-vs-codex", "content_html": "48 hours ago, OpenAI released Codex, their lightweight coding agent that looks like it does a lot of the same stuff that Claude Code does.
\nBecause Task Demon uses coding tools like Claude Code to create the detailed plan documents it writes, we were excited to immediately add support for Codex in Task Demon. With v0.7.0 of the Task Demon CLI agent you can now just set the tool option to codex in your .taskdemon/config.yml file and the agent will start using Codex. Simple, huh?
Ok so let's see how these two tools compare.
\nIs it silly to compare the two? Codex is brand new, and Claude Code is relatively mature at all of 2 months old. Codex will evolve rapidly, and we've only had a few hours to play with it, so these initial observations are just that - initial observations.
\nMost of what Task Demon does is to very efficiently create plan documents for all your coding tasks, so that's what we focused our testing on. We gave each tool a set of 5 planning tasks to perform, and measured how long it took as well as the quality of the output.
\nThe results are of course subjective, and the landscape will change rapidly, but here's a 6 minute video of what we found:
\nFor those who want to read rather than watch, keep going:
\nTask Demon primarily uses coding tools to generate plan documents for some task, so we focused our testing on that. We gave each tool 5 planning tasks to perform, and measured how long it took as well as the quality of the output.
\nIn the results repo there are 4 folders:
\nWe named them agent1 and agent2 because in a moment we're going to have Claude Code and OpenAI Codex themselves assess which made the better plans, and we don't want to give anything away!
Here's an example of one of the tasks we gave each tool - in this case the changelog task:
\nThis task description was turned into 2 plan documents, along with 4 other tasks. Along with the task description we passed the tool a detailed set of instructions on how to generate the plan. They are a little long to show in their entirety here, but let's take a look at some of what they came up with, starting with Claude's plan:
\nThe plan document the Claude spit out includes a lovely diagram, and numbered tasks, which is cool. It largely went along with the instructions we gave it, and it took about 2 minutes to complete the document.
\nNow let's take a look at Codex's plan - again too long to reproduce in its entirety here but follow the link if you want to see the whole thing:
\nTo me, Codex did not follow the instructions we gave it quite so well. It did attempt to make a diagram, but it doesn't render properly in GitHub, and the tasks aren't numbered. Looking at the other plans output by Codex, only about half of them have numbered tasks, whereas all 5 of the Claude plans use numbered tasks as requested.
\nHowever, the agents themselves have a different perspective, as we will see:
\nOur final step was to ask Claude and Codex which of them had done a better job with the generated plans.
\nWe prepared a detailed PROCESS.md prompt for the tool to follow, and asked them to follow the instructions.
\nHere's what Claude Code said:
\nNote that it actually says \"agent2\" instead of \"Codex\" because we never told it which one was which - here's the full raw output.
\nCodex's verdict was a little more nuanced:
\nYou can see Codex's full verdict here.
\nWe ran these several times and generally got consistent outcomes - Claude Code said Codex's plans were better most of the time, Codex said it was about tied. And to the only human in the loop, Claude's plans look better.
\nMy anecdotal experience using the 2 agents over the last day or so is that Codex is plenty capable, but it also takes a lot longer to do things than Claude Code does, and it tends to be a lot more variable in how long it takes. Sometimes it will generate a plan in 30 seconds, which doesn't seem long enough to generate a good plan, other times it will take 2-3 minutes for the same task. There are a million reasons why this could be the case today and be solved tomorrow.
\nBeyond that, Codex feels a little more timid than Claude Code. After the plan documents were generated, we had both Codex and Claude go and implement some of the plans, reverting the changes between each try. Codex generally took longer to get the task done, but what was notable was that it would often spend the first several minutes just reading code - even when presented with a detailed plan document. It is fascinating to watch these machines make hundreds of tool calls, but watching Codex it was sometimes a little puzzling why it was reading certain files that didn't seem to have much to do with the task.
\nAre we going to keep Codex around? Absolutely. It's already in a great place for a tool that's been released for 48 hours, and the fact that it's open source is a big plus for a lot of folks who want more control over where their code gets LLM'd. Task Demon has first-class support for OpenAI Codex from yesterday, and we're really excited that Claude Code has more competition to contend with.
\nSo install codex now, and give it a spin. Don't expect it to beat Claude Code just yet, but do expect it to keep getting better, fast.
", "url": "https://taskdemon.ai//blog/code-vs-codex", "title": "Code vs Codex", "image": "https://taskdemon.ai/images/posts/2025/Claude-plan-document-tasks.png", "date_modified": "2025-04-18T19:27:09.000Z" }, { "id": "import-export", "content_html": "We're pleased to announce a new feature available today on Task Demon: the ability to easily export and import entire projects. This update was inspired by your feedback and our commitment to data portability, ensuring you're never locked into a single platform.
\nAt Task Demon, we understand your project data is crucial and should be fully under your control. With this new functionality, you can quickly export all your tasks, documents, workflows, chats, and other project data into a single JSON file. Whether you want to keep a backup, move your data between different accounts, or even migrate away from Task Demon entirely, we've made it straightforward and reliable.
\nExporting Data
\nImporting Data
\nWe've included thoughtful touches, like mapping user IDs appropriately during import, to minimize confusion and streamline the process.
\nWe firmly believe that you should be free to use the tools that best fit your needs. By providing seamless export and import options, we're ensuring Task Demon stays true to this belief. Whether you're importing data into Task Demon or exporting your data to explore other solutions, the process is simple, transparent, and designed with your convenience in mind.
\nHead to your project settings page today to explore this new capability. As always, we welcome your feedback and suggestions to continue improving Task Demon.
", "url": "https://taskdemon.ai//blog/import-export", "title": "Export and Import Your Task Demon Projects", "image": "https://taskdemon.ai/images/posts/task-demon-hero.png", "date_modified": "2025-03-28T19:27:09.000Z" }, { "id": "introducing-task-demon", "content_html": "In the ever-evolving landscape of software development, a new paradigm has emerged that we affectionately call \"vibe coding\" — rapidly iterating with AI coding tools, generating solutions in a fraction of the time it once took. While this approach has revolutionized productivity, it often comes with significant tradeoffs in quality, structure, and maintainability.
\nToday, we're thrilled to introduce Task Demon, a comprehensive task management system designed to bring structure, quality, and efficiency to AI-assisted software development while letting you maintain the fast-paced experience you love.
\nThe rise of AI assistants has fundamentally changed how we write code. We can now describe features in natural language and receive full implementations in seconds. This \"vibe coding\" approach feels almost magical — but it comes with hidden costs:
\nTask Demon bridges the gap between the speed of AI-assisted development and the quality of traditional software engineering practices. It's not another coding AI — it's a structured task management system that optimizes how you use the AI tools you already love.
\nAt its core, Task Demon brings a workflow-based approach to AI development:
\nTask Demon starts by gaining comprehensive understanding of your project through automated analysis. This characterization process:
\nWhen a new task arrives (whether created in Task Demon or imported from GitHub), the system automatically:
\nThis structured approach ensures that even complex tasks are broken down into manageable, well-defined steps before a single line of code is written.
\nPerhaps Task Demon's most powerful feature is its ability to generate hyper-specialized prompts for your AI coding tools:
\n<purpose>\n You are an experienced senior software engineer working on a code repository.\n You will be given details about the Project (which represents the repository), including any existing\n high-level documentation about the Project. You are being given a detailed plan for some work that needs to be done, including tasks, files, and instructions.\n The plan has been prepared for you ahead of time by an LLM Architect, and then checked and improved\n by a second Architect. It is likely that the plan is good and you should follow it as closely as possible to completion.\n</purpose>\n<instructions>\n <instruction>Implement all of the tasks in the plan, one by one, in the order specified</instruction>\n <instruction>For each task, ensure that the work has been completed according to the instructions in the plan</instruction>\n ...\n</instructions>\nThese prompts contain:
\nThe result? Claude Code and other AI tools produce dramatically better code that aligns with your project's architecture and standards.
\nWe take the privacy and security of your code extremely seriously. That's why Task Demon follows a privacy-first design:
\nThe CLI agent is a simple NPM package that runs locally, analyzes your code, and communicates only necessary information back to Task Demon. This approach ensures your sensitive code never leaves your machine.
\nTask Demon doesn't want to be the source of truth for your project tickets — it integrates deeply with GitHub to enhance your existing workflow while adding AI-powered capabilities. Our GitHub integration works on multiple levels:
\nWhile GitHub is our first integration, our platform is designed to work with other ticketing systems in the future, including Jira, GitLab, and Linear.
\nTask Demon provides specialized AI agents for different aspects of your workflow:
\nThese specialized agents work together to provide a seamless, intelligent workflow that adapts to your project's specific needs.
\nAfter extensive testing of AI coding workflows, we've discovered that the key to success isn't just having better AI models — it's having better processes. Task Demon embodies software engineering best practices while embracing the speed and capabilities of modern AI:
\nTask Demon is now available for teams of all sizes. Getting started is simple:
\nnpm install -g @taskdemonai/clitd logintd agent:startWithin minutes, you'll have a powerful AI task management system that enhances every aspect of your development workflow.
\nWe believe Task Demon represents the future of AI-assisted development — where AI tools are guided by structured processes, deep project understanding, and software engineering best practices.
\nOver the coming months, we'll be adding:
\nJoin us on this journey as we redefine what's possible with AI-assisted development. Get started today and experience the perfect balance of speed and quality.
", "url": "https://taskdemon.ai//blog/introducing-task-demon", "title": "Introducing Task Demon: Structured AI Task Management for Modern Development", "image": "https://taskdemon.ai/images/posts/task-demon-hero.png", "date_modified": "2025-03-23T19:27:09.000Z" } ] }