Scalable agents working on your Github Issues, built on top of the SOTA agent on swebench-verified From the creators of Aide the AI native editor

London
Pinned Tweet
This one is for the devs building frontend applications 🥁🥁🥁 Aide now ships with a browser and React Devtools built-in! Attach screenshots & code just by clicking around your app 📸 ...and, the agent *looks* at the browser every few steps to visually verify its work! 👀
7
2
72
17,085
Hey all! It is with a very heavy heart that we have made the decision to sunset Aide. Users with an active subscription will receive a refund within the next 5-7 days, and your access to the CodeStory provider will work till the end of the month. Access for all free users is being revoked immediately. The editor and sidecar are both Open Source, and can be run on your own with API keys indefinitely. We made great strides this past year, shipping an editor with a small team of 4, getting to the top of swebench twice, churning through 50B tokens a week. But, at the same time, we had a lot of bugs we couldn't keep up with, got out-executed by teams with a lot more talent and capital and importantly, failed at converting our wins into a distinct and sustainably growing product. We are tremendously grateful for everyone who believed in our vision, stuck with us during the early days, during the times the editor was a mess, and the times when your trust in our team and the product paid off. As a company, CodeStory is going nowhere yet and we have new avenues that we are exploring - more on which we will be sharing soon. A huge thanks to everyone again
19
2
111
9,541
Say hello to agents in your editor! 🤖 🤝 🤖 We are taking out first step towards putting agents right in your editor. Unlike other platforms, its not just a single agent but many of them working, collaborating together to get the task done. You can see the swarm of agents working together to explain a very "long workflow answer"
2
7
33
6,519
SWE-agent right in your editor. We are launching a new extension which provides the best of our agents in your editor of choice. It can: - read directory files - run terminal commands - edit files - and more... Here is our SWE-agent editing the website itself 😉
3
1
27
1,415
We achieved State-of-the-Art on SWE Bench Lite with 40.34% accepted solutions!
Excited to share that our team has cooked up a multi-agent coding framework and setting the new State-of-the-Art on SWE-Bench-Lite with 40.3% accepted solutions! Very soon this framework will work right in your editor with the developer working along with agent(S)
1
1
25
1,692
Aide is moving from fully open access to a paid subscription model. Starting today, we’ll offer a $20/month plan with unlimited agentic features and early access to top models
5
2
25
3,242
📸 Sidecar v0.1.14 - Early alpha support for o1-mini to guide sonnet3.5 We are excited to announce Agentic reasoning. We heard your feedback about our agents going off-track and doing too much, we have an objective answer to it
6
1
21
2,939
Several agent improvements landing on Aide: - reduced doom looping, sonnet3.5 gets stuck less often - improved tool usage (mostly because of parsing improvements) - new file creation bugs no more! - terminal use from correct working directory
2
1
21
1,893
Aide 1.0 is now available, and live on @ProductHunt ✨ Aide is an AI-native IDE with privacy at its core — and is the world's first editor with built-in support for @AnthropicAI's Claude Opus, Sonnet and Haiku! 🧨 These models can be used for Inline code completion, chat and inline chat, all of which are built-in to the editor. Opus performs better than GPT-4 in coding tasks, and Sonnet offers the perfect mix of speed and reliability required for AI-assisted coding. You can either use the models with our paid plans, or bring your API key 😌 Upvote us on PH, and try the editor today! We're also hand onboarding 100 developers to the IDE, and taking AI-native feature requests that you're looking for in your IDE. If you're interested, just sign up from the pricing page. We've never relished more in the joy of programming.
3
1
21
6,277
you wanted support for any and all models, its here Testing out the sonnet via @openrouter Cmd+Shift+P: Model Selection (JSON) to any model!
1
4
20
1,857
> be you > get $20 subscription > get unlimited and full context access to o1-mini, sonnet3.5(new), deepseek v3, gemini 1.5 pro > profit?
2
19
2,170
📸 New release for editor and sidecar - Aide now supports .aiderules use this similar to how .cursorrules works and set custom instructions for the edit and chat mode to use - The agent does not keep editing the same file agian and again, no more "formatting issues" - "Rage Shake" you can now report errors directly to us developers by doing a rage shake when the chat/edit/agent does not work and we capture all the required telemetry about it
4
1
19
2,396
AI should not be living in the cloud, it should be in your basement. We believe in complete data ownership and local-first experiences for AI coding tools. Rather than a centralised API, each organisation and individual developer will soon have access to models that they run at their end, which can be customised to their needs and offer complete data security. We are taking our first steps in embracing this future with Aide, by adding support for two Open Source models — Mixtral and Mistral from @MistralAI! You can connect to the supported models on Aide today through @OpenAI, @togethercompute, @Azure and… locally @ollama! Ollama, and models running locally, is special because AI can assist us with programming on every keystroke. But today because each call goes to the network, there is latency and cost that adds up quickly. As local LLMs become more capable to run on developer machines, we can unlock plenty of new experiences beyond code autocomplete and chat.
This year it looks like >4 companies will reach GPT-4 level models Given its lead, one could guess OpenAI should be at GPT-4.5 or 5 This should reshape some segments of both AI world, as well as open up the next set of industries to transformation by GenAI
3
4
15
3,944
all in a day's work
CodeStory agent is now SOTA on swebench-verified with 62.2% resolution rate. We did this by scaling our agent on test time inference and re-learning the bitter lesson. Sonnet3.5(new) was the only LLM we used for this run
3
3
17
1,353
Say hello 👋 to your omni-present AI agent Our AI agent is present with you every step of the way, start talking to it today with @ context symbols and / commands for easier workflows. `Shift+Shift` to the agent
1
14
2,241
105M tokens/day @openrouter
1
5
14
1,285
We are improving the UX for probing: - We now show the results of the probe directly in the editor - Decorations are improved so you can see where the agent decided to click and the symbol to follow
1
2
14
1,194
you can also vibe code from your phone using agent farm
watch me speed run adding sonnet3.7 using sonnet3.5 and making sure the code compiles 😇 all from my phone which was at 4% battery (these are tiring times)
1
15
2,927
📸 Sidecar 0.1.12 is out now! This release includes: - Improved retry logic, this shows significant boost in our evals and reduces the doom loop with `Thinking ...` scenarios
2
12
3,315
Major update for sidecar: - we updated the sidecar version to v0.1.35 which fixes the annoying "LLM did not respond in the correct format" - along with several updates to our backend, you should not see this error anymore restart and update sidecar today to profit!
2
2
13
1,380
@deepseek_ai has one of the best coding models out there! you can use it with @togethercompute today with Aide!
@deepseek_ai coder 33B is one of the most unique models which supports almost all of the in editor workflows.. - 〰️inline chat - 🪟sidebar chat - 👻 ghost text look at that speed with inline chat!
1
3
11
4,146
Replying to @tunahorse21 @skcd42
fixed the setting! you can DM the account directly too
4
69
Long context 🤝 Gemini-Pro-1.5-Flash Try out full codebase search with in-context learning and search powered by Gemini-flash models
Long context is now available on Aide! We now use Google's Gemini-Pro-1.5-Flash for performing long context search and starting the agents. You do not need to be at the right file or location and can start the work from anywhere.
1
11
1,175
We have onboarded close to 200 repos to CodeStory! ✨🚀 Developers using the IDE are leveraging the built-in AI agent to make changes across their codebase, chatting with their code, semantic search and AI-enhanced change tracking. Try us out at codestory.ai ❤️
10
1,489
Say hi to Aide ✨ your AI first editor Try out Aide for free now (on your Macbook) here: github.com/codestoryai/binar… With Aide you can do: - AI agents which can do repo wide edits - AI which takes care of keeping your commits - Semantic search over your codebase
8
1,326
thank you for the callout!
We just updated the SWE-bench leaderboard with *14* new submissions! Congrats to @blackboxai and @aide_dev on setting new SoTA scores on Lite & Verified! Congrats to @Google on your first submission. Credit: @jyangballin, @_carlosejimenez, @KLieret who maintain SWE-bench/SWE-agent
2
10
1,110
🚨 We had permission issues because of an unzip library change. This is leading to "Failing to start sidecar" issue We have identified the issue and releasing a fix
3
10
1,697
Introducing CodeStory Chat and slash / commands ✨ - With CodeStory chat you no longer have to jump between your browser and editor to ask simple questions, simply select the code and let CodeStory do its magic 💬 - invoke Aide with /agent to make bigger changes 🤖
1
1
9
688
Coding at the speed of thought
Coding at the speed of thought. Just spitball, and LLM codes. Powered by agents. Background refactors coming next.
2
9
861
Iterate and build
Code at the speed of thought The agents are writing code in real time, no speedups! Instead of coming up with the proper prompt, you can think through the changes you are about to make ITERATIVELY and the agents take care of writing out the code.
1
8
1,522
We are deprecating our Github PR review bot 🥹 This will allow us to focus on the editor and ship more features for it. We are thankful 🙏 for the folks who beta-tested the PR bot and all the infra we built for it will be shipping to the Aide editor. We plan on bringing it back at some point in the future 🔜
8
2,074
For Vite + React projects (including Remix!), the "provide context with React inspector" tool extracts the exact code range of the component that maps to the DOM node. First-class support for Next.js is ships next week too ✨
3
9
381
Replying to @withneo
👋
7
598
We have memory support 🧠! You can now ask followup questions to any of the /search /explain workflows and the agent will understand what you are talking about. On top of this, you can use our agent as a general coding assistant all while staying inside the editor :)
2
7
1,091
Search for workflows and not just code pointers! We overhauled our search experience so the agent can reason across multiple files and provide answers which require reasoning. Mimicking what humans do by reading files, keeping context and creating a story about the changes!
1
6
1,341
We are now inline in your editor! ♨️ Want a quick documentation generator for your function or class, call on Aide by clicking `Cmd + I` and you can ask Aide to generate documentation inline. We are going to add more skills to the in-line agent very soon, stay tuned!
2
2
6
1,614
We plan on digging deeper into runtime information from the framework as well as the browser - until we have a perfect trifecta of the editor, the browser and AI agents, coming together to offer the most powerful developer experience for frontend engineers.
1
5
903
Providing screenshot as a tool for the agent has massively increased the visual accuracy of the code changes that it makes. When providing Figma design screenshots, we've seen the agent be very precise about the changes it makes and iterate until perfection 🔎
2
8
315
Paid tier ($20/mo): - Unlimited sonnet3.5(new) - o1 (30 requests/month) - Gemini-2.0-flash and the other latest models as they drop - priority 24/7 support
1
7
595
Lot of new fixes for the tab-autocomplete
1
1
6
2,035
🫡
Contribute to Open Source using AI 🚀 @aide_dev is an open-source AI-native IDE that leverages Agentic workflows and external tools to understand large codebases and dependency graphs, enabling it to generate more accurate responses. AIDE was benchmarked against SWE-Bench Lite, achieving SOTA performance by resolving 43% of the issues. It also serves as an excellent free alternative to Cursor AI. (UPDATE: CodeStory agent is now SOTA on swebench-verified with a 62.2% resolution rate.) Kudos to @skcd42, @ghostwriternr, and team.
2
6
669
To enable tick the following setting: Enable Agent Reasoning in your User Settings. Its early alpha so please share any and all Rage Shakes our way! ❤️‍🔥
1
5
1,011
pushed an update to the agent farm which improves performance when running terminal commands! if you were experience the agent not iterating properly on `cargo` / `pytest` / `jest` errors, the agent should focus on these even more and iterate better on the errors
1
7
4,027
Quality of life updates: You can now use #openFiles in the editor to add them to the chat context automagically
1
6
354
👻 Tab-Autocomplete away 🔜 One of the most high throughput and useful AI features to date for coding is Tab-autocomplete, it helps you stay in flow and copilot just understands it 💡 We are adding native support for Tab-autocomplete using the following LLMs: - Codellama powered by @AIatMeta - Deepseek coder instruct models @deepseek_ai For the providers you can pick you favorite: - use @ollama - @togethercompute (they have one of the fastest endpoints) - or your own vllm hosted LLMs (from the code llama or the deep seek variants)
1
1
5
1,516
✨ New builds for the editor are out, including the latest sidecar! ✨ This release includes: 1. Improvements to the agentic mode to reduce the doom loops, and increase the overall speed
1
6
1,306
look ma I got terminal powers
terminal coming soon ;)
1
6
639
🎙️🎙️
codestoryai / sidecar: Sidecar is the AI brains for the Aide editor and works alongside it, locally on your machine ★229 github.com/codestoryai/sidec…
1
6
520
CodeStory is an AI 10x engineer helping developers with day-to-day tasks that go beyond writing code 🚀 You can find our extension here: marketplace.visualstudio.com… We currently support the following features: - 🔍 Look up code using the context you remember in natural language
1
5
3,325
BYOK (Bring your own key) You can use your own OpenAI key with CodeStory, this allows you to share the cost and provides better performance from the agent and the chat. By default we take on the cost, and in future we will allow you to bring your own LLMs to power the editor
1
5
664
✨ New Builds ♨️ We are shipping alpha version of our editor on all major platforms. Join us on our discord to catch the latest developments: discord.gg/Cv7pShdgrD Mac M chip: github.com/codestoryai/binar… Mac Intel chip: github.com/codestoryai/binar… Windows x86: github.com/codestoryai/binar… and github.com/codestoryai/binar… Windows arm64 github.com/codestoryai/binar… Linux: github.com/codestoryai/binar… You can find a list to all binaries here: github.com/codestoryai/binar… We are excited for you guys to try out our editor 🖥️! We use the new sidecar binary everywhere for lots of new features: - our semantic search - our agent based search - autonomous search Join us as we keep shipping 🚢
2
4
556
The agent now supports lexical search 🔎 We replaced the wonky implementation with github.com/quickwit-oss/tant… making the search faster and more powerful. On top of this, we now use statistics from git log so the results are more relevant to your query.
1
4
880
Starting today, we want to make Aide the best editor for developers to work and collaborate along with multiple agents. This is a bold new vision and future we are working towards and we are super excited to share our early version with everyone.
4
362
I can now stream 🎏 We added streaming output support from the backend so you get the response as quickly as the LLM can generate it. Along with accurate markdown support this will help you get your answers that much quicker
1
5
1,171
How we got rid of `// Rest of the code ..` using a Mistral 7B finetune One of the challenges when working in a coding editor is that the generated code is often incomplete with parts of the code commented out with `// Rest of the code ..` or `// ..`
1
5
631
v0 of our fix using Aide is rolling out You can use this to inline errors (it works pretty well if the error is contained in the same file, we are working on providing multi-file fixes)
5
454
Smoother scrolling while streaming output 📜 Before
2
5
682
feedback loops are all you need
5
650
Ship anything with Aide today at aide.dev 🚀
5
809
We hit our rate limits on OpenAI, hang tight as we get this issue resolved on our end!
2
1
5
283
We now stream the responses for the chat view 🏄‍♂️ This also has stability wins as the streaming endpoints for most LLMs (OpenAI and Llama2) are more stable in general compared to non-streaming endpoints.
1
4
460
We still offer a free tier with 50 free invocations/month. This keeps Aide accessible while we invest in advanced features & support for those who depend on us most.
1
3
657
🚀
annoying ';' SOTA-SWE got your back! from prompt to fix in under a minute in a production codebase. The future is now!
4
851
tune in to lear more about the state of AI codegen!
Gonna start the new year with more AI code-gen related content. Tune in to learn more about: - bitter lesson - why reasoning models are a big big deal - "fix all bugs in my codebase" nitter.app/i/spaces/1vOGwryRpAbKB
1
4
918
Replying to @manuhortet
thank you! we will be back soon, the shipping never stops!
4
1,590
And the developer stays in the loop for all the actions performed by the agents, fully able to dive in and understand what the agent is up to.
1
3
448
Replying to @tunahorse21
DM?
1
2
63
We will be adding more functionality to reference code symbols and add this to the inline-agent flow very 🔜
2
75
To support the community of developers building/fine-tuning LLMs, we provide full access to the SQLite DB, where we store the prompts and responses for every LLM call locally. This data can be used for further fine-tuning and building extensions. Our long-term plan is to provide better and more powerful abstractions for developers building integrations with models of their choice. In that direction, we are starting by Open Sourcing the prompts we are using in the IDE for each model. We have tuned and tested each of these at our end to get the best result from each model. github.com/codestoryai/promp…
1
4
179
We shipped cloud backed explanations and semantic search. Instead of indexing locally we provide ☁️ backed understanding of your code. We index the repo on every commit to main, and provide value on your editor the moment you open our Extension!
1
4
203
The gifts keep on coming 🎁 Try out DeepSeek33B for inline-autocomplete and chat today on CodeStory provider without any cost! @deepseek_ai has one of the best coding models and it would be sad to deprive our community of that. See the video below on how to change the copilot model 👇
2
1
3
937
A single agent is scoped to work on a code-symbol (classes/functions/enums.. etc) and talks to other agents (code-symbols) via natural language. How does the agent even know which code-symbol to talk to? Well since we have access to the editor, the agent can literally perform clicks
1
3
373
Semantic search now uses the new sidecar binary, the results are more relevant and more correct. Try it out by hitting the 🪄 on top of the search bar
4
638
red squiggly lines... go away
LLM cleans up after itself [GONE WRONG] Don't waste time on AI errors - they will fix themselves. Acknowledgements: - LSP - Recent edits - Sonnet 3.5 (r.i.p.)
4
478
Aide ✨ now supports python! With support for python, you get a powerful AI agent which can do repo wide changes and all the goodness of CodeStory right in your editor. Give it a try today!
1
4
205
🔍 cite your sources! We now 🖇️ directly to the diff page, allowing easier navigation to the diff page and in context of the answer. Soon codestory will link to all its sources for generating answers!
1
3
550
This week has been a fun time for the Aide! 🎉 To end the week on a high note, you can use @OpenAI 's GPT4-Turbo from the CodeStory provider for free! You can put up-to 20k tokens (or 26k words approx) and custom system instructions!
1
3
593
Replying to @tunahorse21
damn my bad, DM @skcd42 for anything and everything
2
2
100
Typescript on ☁️ We support typescript on our cloud based offering, allowing users to get the latest code symbol graph on of their repository without having to index locally! We power both our 🔍 semantic search and 🤯explain functionality on our cloud offering!
3
269
Integrate incomplete code snippets from LLMs straight into your editor - no more '// rest of the code...' or '// ...' placeholders. 🛠️ We use smaller LLMs to edit directly in your file, bridging 🌉 the gap between the editor and the chat window.
3
1,291
Create better commit messages capturing the "what" of the change automagically 💫 and focus on the "why"! @codestoryAI helps you commit your changes on the same view where changes are grouped. We are rethinking the whole workflow from writing commits to reviewing! 🚢
3
343
Read about how we got the in-line agent to generate code and stream it right in the editor! codestory.ai/blog/stream
1
3
1,741
We celebrate small wins :') We recently crossed 500+ repositories who have tried out Aide and are getting value out of it 🎉
2
369
Replying to @X0Radi
too slow, lack of internal cot destroys the ux, waiting for gemini 2.0 thinking to drop. Might add it depending on how the experiment goes
3
142
print("The magic begins on port " + port)
1
3
108
Try out the IDE and let us know which models and platforms you'd like to see us support! And as always, we're open to hearing your feedback on our Discord. discord.gg/JEjtaVwyeX@skcd42 and @ghostwriternr ♥️
3
200
You can now /explain a function or block of code properly. It also hyperlinks properly to the internal functions which are present on the file automagically ✨ Give it a try by downloading our latest release from 👇 github.com/codestoryai/binar…
2
492
We know this may disappoint some of you. We don’t take this lightly. Our community matters. Share your thoughts and questions—DM us or reach out over the next few days. After the transition, supporters get priority feature requests, faster replies, and special Discord badges. Your understanding helps us make AIDE the best AI agent.
2
721
Along with the UX improvements, we also: - improved the models we are using for go-to-definition - reduced the latency of the overall time it takes for the agents to reply back
3
322
We now support support for all languages using LSP + tree-sitter. It's been in the making since last week and we are GA for GoLang and other languages.
1
3
498
Bunch of fixes landing 🔥 - We fixed the editor on Linux distros, we had an issue with our extension packaging github.com/codestoryai/binar… - You can now use your own OpenAI key again! We lost this when moving to sidecar and now its back
3
165
And the fresh new landing page on the internet 🧑‍🍳 from our chefs @skcd42 and @ghostwriternr. aide.dev
1
3
434
Replying to @erikvbos
thank you for the support!
1
754
Detailed analysis of how we performed on swe-bench-lite and some gotchas!
We are sharing a more detailed breakdown for swe-bench-lite evaluation results and some gotchas which we found. To start with we are scoring 40.3% (the real score is a bit higher than that, but some tests were failing but we got a green tick for it) Now for the gotchas
1
3
687
We are excited about a new suite of feature requests that our users and we ourselves, as developers, are excited to build this week! Regardless of whether you use our editor, let us know what you'd like to see in an AI-first IDE! 🧵
1
2
448
> Our customers are using codestory.ai to quickly understand repositories! CodeStory.ai onboarded @MosaicML's composer repository (3.2k code symbols!) so users can try out codestory with just the extension. Its 3 simple steps listed below 👇
1
1
2
409
always on, humming along
Replying to @siddish_ @aide_dev
its taking 5-10 minutes for each PR, silently works in background, even says beep boop when it can't iterate any further 😂
3
3,250