Have something to say?
Email/Codes to sign in aren't working
I am trying to create an account. I can’t login via email or with it sending me a code. When I try to login via email, it directs me to use the code login access. When I request login access via the code it gives me this response: “ That code isn't valid or has expired. Request a new one.” Help with this would be appreciated.
Can`t sugn up
I tried to sign up on your website using both OAuth and the login/password approach. I received an email with a 6-digit code. However, when I try to sign in, I am redirected to the page https://build.chipp.ai/#/login?error=payment_required
Chip says I’m out of usage credits
Chipp says I’m out of usage and thinks I’m on the free plan but I’m actually on the $30 per month plan. I’m using chip on a browser on my phone.
Notification Badge is Clipped by Parent Container
The unread notification badge (blue dot) on the chatbot launcher is being clipped by the parent container's bounding box, which results in a truncated/cut-off appearance. Could you please fix this so the notification dot should renders completely intact, without any clipping? Something like this: Thank you!
My apps are missing
Hi there, I just logged in and noticed the updates (great) however all my apps are missing from my dashboard, but they’re active for users to still create accounts from existing links. This is concerning, especially because the free tokens are set at 100 when i originally had them at 6. I can’t go in and fix it cause the app is missing.
MCP error: add_url_source via MCP never triggers crawler (sources stuck PROCESSING)
Date: 2026-04-05 MCP Server: Chipp (build.chipp.ai/mcp, streamable-http transport) MCP Client: Claude Code v2.1.92 (Anthropic CLI, macOS) Steps to reproduce Call add_url_source via MCP to add a URL-based knowledge source: { "appId": 10041051, "url": "https://portal-conneva.vercel.app/api/mosaic/knowledge/index.html", "crawlAllLinks": true } Result Source is created successfully and returns a source ID (e.g., dd18f40d-5de9-46a1-a1c8-2a41a9686237). However, the source remains in PROCESSING status indefinitely — the crawler never picks it up. Calling get_crawl_progress on the source returns: { "error": { "code": "INTERNAL_ERROR", "message": "Failed to connect before the deadline" } } Expected Source transitions through PENDING → PROCESSING → COMPLETED as the crawler fetches and indexes the content. This is the behavior observed when adding the same URL via the Chipp dashboard. Root cause evidence Dashboard-added sources crawl fine. The same URL was added manually through the build.chipp.ai dashboard and crawled to COMPLETED status successfully. Two individual topic URLs were also added via dashboard and indexed correctly. MCP-created sources never start crawling. Multiple sources created via add_url_source MCP tool remain stuck at PROCESSING. No crawl activity is observed. The crawler itself is healthy. Dashboard-initiated crawls complete normally, confirming the crawler service is running. The issue is that MCP-created sources are not being enqueued for crawling. get_crawl_progress also fails for MCP-created sources with a "Failed to connect before the deadline" error, suggesting the crawl was never initiated (no Temporal workflow to query). Tested multiple times across different sessions and URLs — consistently reproducible. Workaround Add sources manually via the build.chipp.ai dashboard instead of the MCP add_url_source tool. Environment MCP Server: Chipp (build.chipp.ai/mcp, streamable-http, OAuth 2.0 + PKCE) MCP Client: Claude Code v2.1.92 (Anthropic CLI, macOS darwin) All other MCP tools (app management, list_knowledge_sources, delete_knowledge_source, custom actions, system prompt, branding) function normally for this workspace and app.
MCP bug: search_documentation fails with Postgres connection error
Date: 2026-03-22 MCP Server: Chipp (build.chipp.ai/mcp, streamable-http transport) MCP Client: Claude Code (Anthropic CLI, macOS) Steps to reproduce Call search_documentation with any query: { "query": "add_document_source file upload base64" } Result { "error": { "code": "INTERNAL_ERROR", "message": "connect ECONNREFUSED 127.0.0.1:5432" } } Expected Returns relevant documentation excerpts. Root cause evidence The error connect ECONNREFUSED 127.0.0.1:5432 indicates the documentation search handler is attempting to connect to a PostgreSQL instance on localhost:5432. This suggests the docs search feature uses a local or sidecar Postgres database that is not running in the MCP server's deployed environment. Other MCP tools that use a different database connection (app CRUD, knowledge sources, custom actions) all work correctly, so this appears to be a database configuration issue specific to the documentation search feature. Tested twice in the same session — consistently fails with the same error. Environment MCP Server: Chipp (build.chipp.ai/mcp, streamable-http, OAuth 2.0 + PKCE) MCP Client: Claude Code (Anthropic CLI, macOS darwin) All other MCP tools (app management, knowledge sources, custom actions, system prompt, branding) function normally for this workspace and app.
MCP error: add_document_source fails with GCS credential error
Date: 2026-03-22 MCP Server: Chipp (build.chipp.ai/mcp, streamable-http transport) MCP Client: Claude Code (Anthropic CLI, macOS) Steps to reproduce Call MCP add_document_source tool with any file. Minimal example: { "appId": 10041051, "files": [ { "fileName": "test-topic.md", "content": "IyBUZXN0IFRvcGljCgpUaGlzIGlzIGEgdGVzdCBkb2N1bWVudCBmb3Ig dGhlIGtub3dsZWRnZSBiYXNlLg==" } ] } The base64 content decodes to: Test Topic This is a test document for the knowledge base. Result { "error": { "code": "INTERNAL_ERROR", "message": "The incoming JSON object does not contain a client_email field" } } Expected File uploads and begins processing (PENDING → PROCESSING → COMPLETED). Root cause evidence The error message "The incoming JSON object does not contain a client_email field" is produced by the Google Cloud Node.js client library (google-auth-library) when parsing a service account key file. It is thrown when the GoogleAuth constructor receives a credentials object missing the required client_email property. This suggests the add_document_source code path initializes a Google Cloud Storage (GCS) client to upload the decoded file, but the service account JSON key is either missing, malformed, or not configured in the MCP server's runtime environment. Evidence that other knowledge source tools work (same app, same session) The failure is isolated to the file upload code path. All other MCP tools — including other knowledge source tools (add_url_source, list_knowledge_sources, get_knowledge_source, delete_knowledge_source, refresh_knowledge_source) — work correctly for the same app and workspace. Environment MCP Server: Chipp (build.chipp.ai/mcp, streamable-http, OAuth 2.0 + PKCE) MCP Client: Claude Code (Anthropic CLI, macOS darwin)
AI model mismatch
LLMs do not match corresponding models. ChatGPT 5 latest is actually Gemini 2.5 pro, Gemini 2.5 pro is actually Gemini Pro 3 and so on.
Empty or incorrect weblinks
ChatGPT 5 cites references that do not exist. Also web links are often empty or lead to ‘page not found’
User reports crashes
Here is what the users reported: “I’ve had this problem a few times this week—I can’t open a chat in the app anymore; it just keeps loading until it crashes… At one point, I was able to open it again and use it without any issues. Now it’s completely frozen again, both on my phone and my laptop. This is really frustrating for me because I can’t make any progress on my project.” How can we help him asap?
Image Generation Unstable
I have found image generation to continue to become more unstable. Can you check out the biblepics.faithbot.io chipp and see what’s broken?
User reports issue with file download
The users asked the app to summarize the issue, this is what it said: Chronological sequence of events: First attempt (PDF): At the user's request, I created a PDF file containing a summary of the exercises. When attempting to open the generated link, the user received an error message. Second attempt (PDF): Assuming it was an isolated error, I created a completely new PDF file. The result was identical; this link also led to the same error message. Third attempt (Word document): As an alternative, we suggested a Word document (.docx). I created the file, but unfortunately the exact same error occurred here as well. The user was unable to download the file. Detailed error description: The user was kind enough to send a screenshot of the error message. It is an XML error page storage.googleapis.com with the following content: Code: AccessDenied Message: Access denied. Details: Anonymous caller does not have storage.objects.get access to the Google Cloud Storage object. Permission ‘storage.objects.get’ denied on resource (or it may not exist). Analysis and interpretation of the error: Essentially, this means that the files I created on Google Cloud Storage are not correctly shared for anonymous (public) download. This appears to be a general permission issue in the system configuration that prevents users from accessing the generated files.
App produces Bad Request with Gemini 2.5 Pro 🚨
My app suddenly doesn't work anymore. It says bad request. It's running on Gemini 2.5 Pro. When I change to GPT it works. I have to change to a more expensive model to keep similar quality, so my usage cost increases due to technical errors I cannot influence
Search in Chatlogs doesn't work
I want to see the chatlogs of a specific user but when I type in the email into the search bar nothing happens.
User reports with file download
A paying customer complained about broken file downloads. I check the affected chat and it seems indeed that the provided files cannot be downloaded. Also changing the format didn’t help. The affected chat is ID: 17ef026d-1548-4861-81d4-42196ccbd3aa To me, the links to the file downloads seem legit but they don’t start. It’s frustrating, since the cost for Chipp increased massively due to the usage-based pricing. While I think it’s absolutely fair that, we as Chipp users carry these, but then I expect it to work properly. Certainly, these messages with broken file downloads count against my token usage and I am not happy to pay for token consumption that doesn’t benefit my end users at all but rather increases the risk they churn. Here is what the user said: “I am contacting you because I am repeatedly encountering a technical problem when creating a file with the virtual employee Seedy [the name of the app]. I have been trying to create an absence planner for my team for some time now. Seedy has already generated several download links for HTML and Word files for me. Unfortunately, every time I click on one of these links, I get an error message and the download does not start. We have already tried several solutions, including: Generating several completely new download links. Reducing the file size (from an annual planner to a single month). Changing the file format. Unfortunately, none of these steps were successful. As I need the overview for my team planning, I would be very grateful if you could look into the problem.“
Paying user reports continuous issues with image generation
Hi there, a user complaint about the image generation. He said, at some point it doesn’t generate any images anymore. Sometimes, it even says it has technical issues but sometimes it says it created an image but there’s no file. I told him to use a new chat if an existing one gets very long. He said this worked sometimes but even in a completely new chat it only generates 1-4 images and afterwards it doesn’t provide any files anymore. That’s what he said exactly: “I had the app create a prompt for image generation and then copied it into a new chat to generate the image. When I do this, the app can generate 1-4 images in the same chat, depending on how complex the image is. For images that are supposed to look like photographs of real people, it may well be that after the first image is generated, the app will only generate “empty” images if you want more images from the app afterwards.“ And: “Nowadays, it often happens that the app doesn't generate the images at all. The app shows in the chat, “Look, I did this,” but there is no image. So the app “thinks” it has generated it and presented it to me, but no result appears in the chat.” —— With the significant cost increase due to the usage-based pricing I expect these basic features to work robustly. I understand that Chipp cannot carry the cost of the model usage but such “empty” image creations will certainly count against my usage-based quota and I don’t want to pay for tokens that doesn’t return a proper result.
MCP server with Claude Code Bug
Session ID: 75347bd2-0f71-4da4-bde3-b5490660e4a5 Log file: ~/.claude/projects/-Users-angelaradcliffe/75347bd2-0f71-4da4-bde3-b5490660e4a5.jsonl The error: Every call to mcp__chipp__create_app failed with: INTERNAL_ERROR: Invalid prisma.application.create() invocation: The column botProtectionEnabled does not exist in the current database. &n
Can't delete any agents
I tried to delete the test agents I have, first in bulk, then one by one, I get “failed to delete” and no further instructions.
Misleading translation
In the German translation, there is a misleading translation in the custom context menu. When I want to save my user-defined context, the save button is labeled with “Lesezeichen” which is literally “Bookmark” and confusing in this context. A simple “Speichern” (=save) would be much more intuitive.