Backlog

In Progress

In QA

Completed

Main Roadmap

Is it possible to link multiple Agents together in a single workflow?{{input}} → {{Agent 1}} → {{Agent 1 output}} → {{Agent 2}} → {{Returned output}}

In Review

Multi-Step Agent Workflows

Dave of the North

Hi! I would love to have a Canvas like feature where for example emails are written or to separate the code from the chat. Thanks!

Canvas like feature

Pablo

Like other tools of the same style (keyring for example), could Boltai in the future be able to interact with the content of a web page present in a web browser? I know you can copy the url and provide it to Boltai, but it's less convenient than being on a page, launching the Boltai shortcut and working with the page content.

interaction with browser windows

freeshare

To access my server I need to put in the header a bearer token. I don't see that option. Thanks

Authorization for Remote MCP Server

Michael Horwitz

I would love to have model fusion mode. Basically what it does:Run multiple models side by side, run an analysis, and fuse into the best result.So, for example: I’m looking for the best AI tool for macOS... with XYZ features. Then it queries a few models (which you select), and as an output you get one combined prompt (that is combined by one model) Check this feature in action on openrouter (fusion)<a target="_blank" rel="noopener noreferrer nofollow" class="link" href="https://openrouter.ai/labs/fusion">https://openrouter.ai/labs/fusion</a>It works really well, and i’m starting to use it more and more! would be amazing to have it in BoltAI

Model Fusion mode

David

I would love to have in-line commands and references with autocomplete like GitHub Copilot in VS Code.<ul><li>/ = force or strongly suggest a tool/action (ex: /search, /add_memory /code, /summarize)</li><li>@ = target a tool provider</li><li># = reference artifacts, files, etc.</li></ul>A clean autocomplete mechanism really makes this feature useful, but would probably require tagging each option with identifiers (e.g., ‘openai:web_search’, ‘boltai:web_fetch’).

Add / @ # commands

harrisonfloam

When using an AI proxy for OpenAI, Google, Claude, or custom AI providers, it is necessary to add specific HTTP request headers, which BoltAI currently does not support.For example, Cloudflare AI Gateway with BYOK/Unified Billing required the <code>cf-aig-authorization</code> HTTP header.

Planned

High Priority

Custom HTTP Header for AI Provider

Planxnx

1.0.0 (14)It’s a remote SSE server, I got it connected to Claude and Cursor, but the connection fails in Bolt

MCP Server Connection fails

9brb2czbb9

I swear this same server works on the MacIssue occurs both on ios and iPadOS <img data-width="100%" data-align="left" data-featurebase-content-key="646b16f66b8d963816ca5dc7/post/682dbd12a4bd1fb0676caef6/0196f2a9-ff0f-779a-bba0-a451fb813764/b64u-SU1HXzI1MzQucG5n.png" data-featurebase-content-filename="IMG_2534.png" src="https://646b16f66b8d963816ca5dc7.featurebase-attachments.com/c/post/682dbd12a4bd1fb0676caef6/0196f2a9-ff0f-779a-bba0-a451fb813764/b64u-SU1HXzI1MzQucG5n.png?X-Amz-Expires=3600&amp;X-Amz-Date=20260916T220000Z&amp;X-Amz-Algorithm=AWS4-HMAC-SHA256&amp;X-Amz-Credential=DO801TYC4FCVNNEKURKM%2F20260916%2Ffra1%2Fs3%2Faws4_request&amp;X-Amz-SignedHeaders=host&amp;X-Amz-Signature=dd9e1fa11e19ec0327d89b9435a2bf31c347ccb89f2ab885450f397b85ff3832">

Running into MCP issues.

bharat.rameshwar

Allow referencing or including context from previous chats in a new conversation.What I'd like:<ul><li>Ability to attach/link a previous chat (or multiple chats) as context to a new one</li><li>Ideally searchable chat history so finding the right one is easy</li><li>attach/link selectable or all history from Projects</li></ul>Why it matters: Work often spans multiple sessions — being able to pull in context from a past conversation without copy-pasting saves time and avoids losing important decisions/outcomes.

Cross-Chat References

Hey Daniel,I was thinking about what we should tackle next, and I had a pretty cool idea. You know that Google NotebookLM app, where you can basically upload your PDFs and "talk" to them? We should seriously consider add this as part of BoltAI.Why this is necessary:<ul><li>No More Guesswork: The AI stops "making things up" (hallucinating) because it's tied directly to the user's own truth.</li><li>The Ultimate Personal Assistant: BoltAI becomes a personal expert, a research assistant, and a project manager all in one. It turns a "cool" AI tool into an "can't-live-without-it" part of their workflow.</li></ul>What do you think? CheersMax

RAG feature  / NotebookLM clone within BoltAI

unmatchedintel

It will be great if BoltAi could support <a target="_blank" rel="noopener noreferrer nofollow" class="link" href="http://SKILLS.MD">SKILLS.MD</a> spec: <a target="_blank" rel="noopener noreferrer nofollow" class="link" href="https://agentskills.io/home">https://agentskills.io/home</a>This allows the user to add in prompts or scripts in a folder, and BoltAI will trigger the respective skills automatically when the model detected certain trigger words.

Skills.md

PunyDev

Hi Daniel,I hope you are doing well.While setting up the local MCP servers in BoltAI, I have encountered several issues so far. Primarily, I attempted to set up Obsidian, PyCharm, IntelliJ IDEA, and other MCP servers. I’d say it would significantly improve the user experience if there were a feature like many AI apps have—automatically detecting an MCP server and connecting to it (see ChatGPT with Terminal or PyCharm on macOS). Furthermore, it would be beneficial to have a universal MCP configuration library, where users can try a configuration to get started. If it does not work out, they can start tinkering. Lastly, I’d like to share my experiences setting up said MCP servers. The GUI in BoltAI was not beneficial. I was unable to save the mcp.json file via BoltAI and had to edit it through the terminal while BoltAI was closed. After opening, the MCP server either loaded or did not (depending on whether the configuration was correct). Even if the configuration was correct, I opened the built-in window to add an MCP server, only to attempt to save it (without making any changes), and it would refuse, stating that the configuration was invalid. I got it working in the end; however, I’d say many users would give up on it sooner.If you need any more information, please don't hesitate to let me know.Best, Vašek

MCP server configuration

charmingvenca

Please extend access to PDF files in projects to subfolders. This would allow us, for example, to integrate business conditions at the top level and allow access to underlying customer folders.

extend access to PDF files in projects to subfolders

elbephant GmbH

It would be great if you could add a plugin that supports Tavily web search.What would be even better is that you would provide the possibility to add custom plugins

Plugins

Andy De Meyer

Hello,Any idea why Bolt Ai now constantly <img data-width="100%" data-align="left" alt="" data-featurebase-content-key="646b16f66b8d963816ca5dc7/post/6aa52e4cca96e968ae9b1a7e/01a0953c-c56c-78fa-a022-662a202cc7b5/b64u-Q2xlYW5TaG90IDIwMjYtMDktMTIgYXQgMjAuMzYuMTlAMnguanBn.jpeg" data-featurebase-content-filename="CleanShot 2026-09-12 at 20.36.19@2x.jpg" src="https://646b16f66b8d963816ca5dc7.featurebase-attachments.com/c/post/6aa52e4cca96e968ae9b1a7e/01a0953c-c56c-78fa-a022-662a202cc7b5/b64u-Q2xlYW5TaG90IDIwMjYtMDktMTIgYXQgMjAuMzYuMTlAMnguanBn.jpeg?X-Amz-Expires=3600&amp;X-Amz-Date=20260916T220000Z&amp;X-Amz-Algorithm=AWS4-HMAC-SHA256&amp;X-Amz-Credential=DO801TYC4FCVNNEKURKM%2F20260916%2Ffra1%2Fs3%2Faws4_request&amp;X-Amz-SignedHeaders=host&amp;X-Amz-Signature=48a4cd0774815d97a32bad316025f1c4bdf9415dc5d71c0741b8a8be52b08ac9"><img data-width="100%" data-align="left" alt="" data-featurebase-content-key="646b16f66b8d963816ca5dc7/post/6aa52e4cca96e968ae9b1a7e/01a0953c-98f9-72ab-bcde-00fcb4efcead/b64u-Q2xlYW5TaG90IDIwMjYtMDktMTIgYXQgMjAuNDYuMjhAMnguanBn.jpeg" data-featurebase-content-filename="CleanShot 2026-09-12 at 20.46.28@2x.jpg" src="https://646b16f66b8d963816ca5dc7.featurebase-attachments.com/c/post/6aa52e4cca96e968ae9b1a7e/01a0953c-98f9-72ab-bcde-00fcb4efcead/b64u-Q2xlYW5TaG90IDIwMjYtMDktMTIgYXQgMjAuNDYuMjhAMnguanBn.jpeg?X-Amz-Expires=3600&amp;X-Amz-Date=20260916T220000Z&amp;X-Amz-Algorithm=AWS4-HMAC-SHA256&amp;X-Amz-Credential=DO801TYC4FCVNNEKURKM%2F20260916%2Ffra1%2Fs3%2Faws4_request&amp;X-Amz-SignedHeaders=host&amp;X-Amz-Signature=4f237b62e7f98a318223574cfc288ef96d86c560847acd444d18eb1587f2c724">hangs when using my Ai workflow shortcut for text and spelling correction??CheersStefan

Bolt Hangs

Stephan Fischer

## The ProblemAnthropic just released Claude Opus 4.6 Fast Mode — a research preview that delivers up to 2.5x faster output token generation. To use it, you need to pass two things that BoltAI currently doesn’t support:1. A custom HTTP header: <code>anthropic-beta: fast-mode-2026-02-01</code>1. A custom body parameter: <code>"speed": "fast"</code>The model ID stays the same (`claude-opus-4-6`), so you can’t work around this by just adding a new model. Without these fields, the request goes through as standard Opus 4.6 — no speed boost, and no way to opt in.This isn’t just about Fast Mode. Anthropic (and other providers) increasingly use beta headers and extra body parameters for new features:- <code>anthropic-beta: context-1m-2025-08-07</code> — 1M token context window- <code>anthropic-beta: prompt-caching-2024-07-31</code> — prompt caching- <code>speed: "fast"</code> — fast inference mode- <code>inference_geo: "us"</code> — data residency controlsRight now, none of these are accessible from BoltAI.## The SuggestionIn the Edit Model screen, add two optional fields:- Custom Headers — a key-value input (or raw text field) for extra HTTP headers.- Custom Body Parameters — a key-value input (or JSON field) for additional parameters injected into the request body.### Example ConfigurationFor Opus 4.6 Fast Mode:|Field |Value ||-----------------|--------------------------------------||Model ID |`claude-opus-4-6` ||Custom Header |`anthropic-beta: fast-mode-2026-02-01`||Custom Body Param|`"speed": "fast"` |For 1M context window:|Field |Value ||-------------|---------------------------------------||Model ID |`claude-opus-4-6` ||Custom Header|`anthropic-beta: context-1m-2025-08-07`|## Why It Fits BoltAIBoltAI already gives power users granular control over model parameters like Top-P, Temperature, and Frequency Penalty. Custom headers and body params are the natural next step — they unlock every beta/preview feature from any provider without waiting for BoltAI to add explicit UI support for each one.This is a forward-looking solution: instead of playing catch-up with every new API flag, you give users the tools to configure it themselves on day one.## PriorityHigh — Fast Mode is available now and providers are shipping new beta features at an accelerating pace. Every week without this means users either can’t access new capabilities or have to fall back to curl/scripts.

[Feature Request] Custom API Headers & Body Parameters per Model

Greg

I am self-hosting a LiteLLM instance to proxy all my AI stuff. I managed to add it as a custom provider and fetch the models by prefixing the API key with ‘Bearer ‘ because the LiteLLM logs showed it was missing that.Now that I added it as a provider and select a model to chat with, I run into this error:<img data-width="100%" data-align="left" data-featurebase-content-key="646b16f66b8d963816ca5dc7/post/67fe7ce8e792d217fbf9bb26/01963a16-2274-7c74-837f-660a369b8f9e/b64u-SU1HXzE2NzMuanBlZw.jpeg" data-featurebase-content-filename="IMG_1673.jpeg" src="https://646b16f66b8d963816ca5dc7.featurebase-attachments.com/c/post/67fe7ce8e792d217fbf9bb26/01963a16-2274-7c74-837f-660a369b8f9e/b64u-SU1HXzE2NzMuanBlZw.jpeg?X-Amz-Expires=3600&amp;X-Amz-Date=20260916T220000Z&amp;X-Amz-Algorithm=AWS4-HMAC-SHA256&amp;X-Amz-Credential=DO801TYC4FCVNNEKURKM%2F20260916%2Ffra1%2Fs3%2Faws4_request&amp;X-Amz-SignedHeaders=host&amp;X-Amz-Signature=de29b2c60f26b403eaf37f2513a8dd863c76f0bfce42418cd90f8e47e1416383">In the LiteLLM logs I don’t see any sign of an API call, so it seems to fail somewhere before it hits the endpoint. I’ve tried to add /v1 at the end of it to no effect.

LiteLLM custom provider throwing an error

boltai

It’s important for the user to monitor and track their AI generation costs, and it should be greater if it can estimate cost of the next generation It should have 1. Tokens usage for each generation (input/output and thinking tokens) 2. Current context window usage, such as tokens of current input, total tokens of all previous messages, system-prompt, MCP and Toolings And it would be awesome if it had a dashboard for generation logs.

AI tokens usage monitoring and estimation

It seems like reasoning is not viewable until after the response is complete. I would like the ability to view the reasoning as it streams, especially for slower local models.

Streaming Reasoning Output

Potatoh Salade

Now we have the agent structure, we should be able to set a default agent for a project. The project would then automatically inherit the system prompt, parameters and so on from the agent. It could also be still possible to add project-level config such as files. At the moment, you have to either manually set the agent each chat within the project, or copy and paste the system message and so on from the agent to the project config.

Feature request: Set default agent for projects

Matt

Show conversation cost like in v1

tiyare

The app currently only allows uploading files with sizes &lt;25MB, which is not enough. My pdf files are larger than this limit.

Remove file size restrictions

Chung Ling Lai

It would be great to see full access to folders instead of adding individual files in a project. Most project structures have sub-folders, and it would be great to be able to add the main folder, instead of individual files. I have not spent much time in BoltAI because of this gap.

Full Directory Access for Projects

Nick Veryha - Ranson

I want to be able to "star" or favorite specific AI models in the model selection list (especially for providers like OpenRouter with 100+ models) so that they are pinned or surfaced to the top for quick access. This is similar to the design pattern used in the Cline VS Code extension. It is very annoying to scroll through a huge list to find frequently used models.

Favorite/Star models to pin them to the top of the selection list

Aquamarine Savanna

In this new version, the toolbar no longer shows the chats in the projects. I need to click on the project and access the chats from the project’s dashboard. This is very unhandy and a step backwards in usability compared to V1. I hope it can be solved soon.

Make chats inside projects visible  in toolbar

paulusnp

As a user of the Bolt v2 app, I am getting an error when I try to continue a conversation. The error message is: "Invalid prompt: The messages must be a ModelMessage[]. If you have passed a UIMessage[], you can use convertToModelMessages to convert them." This prevents me from continuing my conversations in the app.

Error in Bolt v2 App: "Invalid prompt" when continuing a conversation

Purple Rainbow

I'm curious about the new implementation for MCP servers. I would like to test both the Obsidian MCP and the one for Logseq as well. They require manual insertion. Is there a JSON configuration available like in version 1?

Plugins are missing custom MCP addition

Nick Cent

I am using BoltAI via a Setapp subscription, which introduces a slight inconvenience regarding model selection. My Setapp plan does not include AI credits, yet BoltAI still detects Setapp as an active model provider, populating a long list of OpenAI models. Whenever I search for "gpt" to find models linked to my own OpenAI API key, the Setapp models appear in the results as well, often taking priority at the top of the list. Currently, there seems to be no option to hide or disable Setapp models, whereas other providers allow you to toggle specific models on or off. Would you consider adding the ability to disable Setapp as a provider, or hide its models from the selector, in a future update?