Had claude help me define the issue I am having for you. Thanks for the awesome app Issue: IAM role has bedrock:InvokeModel but lacks bedrock:ListFoundationModels (enterprise restriction). Other apps like Cline VS code plugin allow manual model ID input (e.g., us.anthropic.claude-sonnet-4-5-20250929-v1:0) to bypass list fetch. OS version: macOS 15.4 (Darwin 25.4.0) BoltAI version: 2.9.2 Setapp user: Yes also a ios beta tester :-) AI provider: AWS Bedrock with Profile Provider (AWS SSO), Region: us-east-1, Profile: bedrock-api-access Steps to reproduce: 1. Open BoltAI → Add API Configuration 2. Select "Amazon Bedrock" as provider 3. Set Credential Source to "Profile Provider" 4. Set Region to "us-east-1" 5. Set Profile Name to "bedrock-api-access" 6. Click Next 7. Error: "Connection failed: Failed to fetch models from AWS Bedrock" 8. Cannot proceed - no option to manually enter model ID Error message: Connection failed: Failed to fetch models from AWS Bedrock Request: Allow manual model ID input option so users with restricted IAM permissions can proceed when festing models fails.

Ryan E. Mitchell about 7 hours ago

It’s annoying having to deselect tens of old models I’ll never use…

Robert M. Münch 5 days ago

2

I didn’t find any way to get rid of it.

Robert M. Münch 5 days ago

2

I was wondering if you’ve thought of maybe adapting BoltAI to support acting as a front-end for OpenCode? OpenCode has a full SDK here that allows you to interact with it entirely programmatically: https://opencode.ai/docs/sdk/ Given how I already use OpenCode, this was turn BoltAI, overnight, into a *Claw alternative. You could let OpenCode handle all of the backend logic for things like skills, slash commands, MCP, workspaces, model preferences, etc, and simply act as a beautiful, polished front-end with realtime streaming, reasoning blocks, tool calls, etc. I have no idea if this even aligns with your views on what you want BoltAI as a product, or if this is of any value/interset to you. Just figured I’d rbing it up. BoltAI is such a polished native front-end client, and OpenCode has a few alternatives right now (such as OpenChamber, Paseo, and a few others) but I don’t think there’s a really solid native mac app (OpenCode’s official desktop app is electron).

Harsha 5 days ago

2

Thank you for this excellent application. If possible, one request: I would like to have a "JSON" option in "AI Agents" – "LLM Parameters" – "Reasoning Effort" settings. For example, to disable reasoning mode - '{"enable_thinking": false}'.

BW2010 5 days ago

1

Was using grok API with previous version of BoltAi, everything was in working order. Updated to new boltai and grok stopped working. Error Bad Request Retry Message. All other APIs are working. Grok API is working fine in Jan.

inspectorgadget 6 days ago

4

Hello BoltAI team, I have two separate questions about BoltAI 2 on macOS. 1) Cloud sync issue I can not activate the Cloud sync ( Although I have 1 GB cloud storage). The issue I encounter is the following one: Type: Sync Data Validation Error Server rejected sync payload due to foreign key integrity failure (23503) Source: insert or update on table "account_settings" violates foreign key constraint "account_settings_title_generation_service_id_fkey" What I need is simply to understand: what is causing this how to fix it safely. I tried turning off the title prompting but it does not fix the issue., I also do not see clear information in the documentation about how Cloud Sync works in BoltAI 2, what exactly is synced, how local data and cloud data interact, and what the correct recovery process is when sync is paused because of invalid local data. So at the moment I do not know how the cloud sync system is supposed to work, or what I should safely do next. 2) Manual backup in BoltAI 2 I also want to manually back up everything in BoltAI 2, including all chats and prompts. In BoltAI 1, this seemed clearer, but the documentation I found appears to refer to BoltAI 1 rather than BoltAI 2. I cannot find clear instructions for BoltAI 2 on how to manually back up all data in a complete and reliable way. Could you please tell me: where all BoltAI 2 data is stored on macOS which folders/files I need to copy for a full manual backup whether this includes chats, settings, prompts, folders, services, and attachments whether there is an official full manual backup procedure for BoltAI 2 equivalent to BoltAI 1 I find the backup feature extremely important. Thank you.

MubiClarke 8 days ago

1

When you import chats from other services, have a light weight model read through them and add relevant things to the memory database

Anthony DiTano 9 days ago

I was totally blocked to use botai now because sync failed. Seems db primer key issue Could help fix the issue. === BoltAI Sync Error Report === Timestamp: 2026-04-01T07:29:04Z SyncEngineError: Sync encountered an issue. Will retry automatically. Full Error: unknown(SQLite error 19: UNIQUE constraint failed: prompts.account_id, prompts.catalog_id - while executing INSERT INTO "prompts" ("id", "account_id", "name", "content", "description", "tags", "is_private", "import_id", "catalog_id", "catalog_version", "publisher_id", "installed_from_catalog_at", "is_user_modified", "is_following_updates", "created_at", "updated_at", "_status", "_changed") VALUES (?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?)) === End Report ===

Ze Tang (Peter) 18 days ago

5

1) Global Search Across All Providers Currently, the search function in the model picker popover only searches models from the selected AI provider. It would be much more convenient if it searched across all configured providers at once. This way, users wouldn’t need to first select a provider before searching. Instead, they could simply type the model name and instantly see matching results across all providers, making the selection process faster and more intuitive. For example, I have multiple providers set up: OpenAI, Google AI, Anthropic, xAI, Perplexity, and OpenRouter. As it stands, I must first choose a provider and then search within it to find a specific model. A global search would allow users to simply type a model's name and see all matching results instantly, which would save time by removing extra clicks and streamlining the workflow. 2) Auto-Focus the Search Box Bonus suggestion: It would also be great if the search box was automatically selected and focused as soon as the model picker is opened, so users can start typing immediately without an extra click on the search box.

Anıl Yuvalı 18 days ago

Another problem which also exists for PDF Pals, is that after a few days running on my Mac, the app switch from other app to BoltAI 2 GUI is slow. There is a significant delay between the action of switch (e.g., by command + Tab) and the BoltAI 2's GUI shows up.

TomHsiung 22 days ago

1

Some kind of minimal sidebar/outline on the right where you can jump back to different headings in chat. Especially useful in mobile where thumb scrolling may take forever.

Finnthehuckstar 22 days ago

2

Maybe I’m blind, but I can’t find the dates of past chats/messages anywhere. Would be nice. Cheers.

philraymond 23 days ago

MacOS 26.3.1 Bolt AI 2.8.5 Build 60 Claude-sonnet-4.6 Drop excel file into chat. Chat shows excel file when clicked on. Claude says there has been no file uploaded

404wanderer 25 days ago

8

It seems the maximum number of tool calls is set to around eight. After that, it appears to stop completely, and a new message becomes necessary. If you are loading multiple datasets via MCP, this can be quite frustrating. It would be great if the maximum number of calls could be configured like the other chat settings and context.

Maximilian Christl 27 days ago

Unfortunately trying to upload a MIDI file only gives me the option to treat it as a text file, which it isn't. Proceeding raises an error hinting to MIME type ‘audio/midi’. This happens for other file types as well. Couldn't you just allow uploading binary files letting the models handling the rest? Same for iOS?

Moritz Zimmer about 1 month ago

1

Hi Daniel, I wanted to report a few issues with MCP servers that I noticed. Authentication seems to be required much more often than with other clients. I can especially compare it with Raycast AI Chat. In Raycast, I never had to log in again to reauthenticate their HTTP MCP. For BoltAI, it is necessary quite regularly. For Notion, it is extreme. You have to log in again after 30 minutes or one hour. It seems like only the access token from the JWT is being used, and once it expires, the refresh token is not used to obtain a new one. For some MCP servers, the authentication workflow doesn’t work at all. One newer example I noticed is TickTick. https://help.ticktick.com/articles/7438129581631995904 It should work, but you simply get an authentication error after logging in and allowing the connection in the browser. Maybe that’s related to the fact that you can’t add URL-only MCPs like TickTick’s in Notion. You have to add at least one slash after the domain so the Add button becomes active and clickable.

Maximilian Christl about 1 month ago

1

Whenever I use the claude api. I recieve an error after a few responses stating I have attached a video here of the full message I used that caused the error. I would really appreciate your help.

patrickbateman about 1 month ago

3

See Cherry Studio for an example of how this works really well Msty does a decent job of this as well. Only being able to branch into a separate chat thread is klutzy.

Dave G about 1 month ago

1

i everyone, I’m experiencing significant latency issues when using Gemini 3.1 within Bolt. Even for very basic queries or simple code adjustments, the model takes an incredibly long time to generate a response—sometimes up to 10 minutes for a single output. This happens consistently, making the development workflow nearly impossible. Is this a known API bottleneck, or are there specific settings I should check to improve the response speed? Any help or insights from the team would be greatly appreciated.

baxenko about 2 months ago