claude-async-logo
Share
Explore

claude-async-logo
Demo: Claude Async Pack

info

About

With web search and extended thinking, Claude LLM regularly takes 5+ minutes to generate a complete, thorough response. The problem is that Coda (with good intention) closes all API connections after 50s. Maybe you’ve seen it before, the dreaded DEADLINE_EXCEEDED error 🤦‍♂️🤯
That’s why I built the Claude Async Pack. It queues prompts for Claude in Supabase, processes them in Vercel, and returns them to Coda via webhook or polling.
⁠
Pack
⁠
⁠
⁠
Demo
⁠
⁠
⁠
Maker Community
⁠
⁠
todo-list

Instructions

Claude API Token: Generate a new
Claude API Token
⁠
and paste it here ​
​
⁠
⁠
⚠️ ^ The text field above is a “personal control”. Here in Coda, it’s only visible by you. But your API key will be unencrypted in the HTTP request.
Add Row: Add a new row
Add row
⁠
⁠
Add Prompt: Enter your prompt for Claude
(optional) Configure advanced parameters
Details
Model (Optional): The Claude model to use
Default: claude-sonnet-4-20250514
Options: claude-sonnet-4-5-20250929 (Latest), claude-opus-4-1-20250805, claude-3-5-sonnet-20241022, etc.
You can enter ANY model string, including beta releases and new models
Known models show accurate pricing; unknown models estimate pricing based on tier
Max Tokens (Optional): Maximum tokens in response
Default: 4096
Range: 1 to 200,000+ (depending on model)
Higher values = longer responses but higher cost
Temperature (Optional): Control randomness/creativity
Default: 1.0
Range: 0.0 (focused, deterministic) to 1.0 (creative, varied)
Note: Ignored when Extended Thinking or JSON Content is enabled
System Prompt (Optional): Set the context or role for Claude
Example: "You are a financial analyst expert"
Example: "Respond in Spanish"
Example: "Use a casual, friendly tone"
JSON Content (Optional): Format response content as valid JSON
Default: false
When true: Response content will be valid JSON that starts with { and ends with }
Automatically adds JSON formatting instructions to system prompt
Sets temperature to 0.2 for consistency
Use when you need structured data output
Include Wrapper (Optional): Include full API response with metadata
Default: false
When false: Returns just the content text
When true: Returns complete API response including usage stats, model info, stop reason
Use true for debugging or when you need token usage details
Extended Thinking (Optional): Enable extended thinking mode
Default: false
When true: Claude shows its reasoning process
Includes thinking blocks in response for transparency
Better for complex analysis and multi-step reasoning
Temperature is not set (Claude decides optimal value)
Thinking Max Tokens (Optional): Budget for thinking tokens
Default: 4096
Only used when Extended Thinking is enabled
Higher values allow more detailed reasoning
Max Searches (Optional): Enable web search with maximum search limit
Leave blank to disable web search
Example: 5 = allow up to 5 web searches
Automatically adds citations as footnotes to response
Citations formatted as: [1] [Title](url)
Best for current events, recent data, or factual verification
Image URLs (Optional): Add images to your prompt
Single image: Paste URL directly or use image column
Multiple images: Use comma-separated URLs or JSON array
For 2+ images from Coda columns: thisrow.[Image Column].FormulaMap(CurrentValue._Merge().ToText().ParseJSON("$.publicUrl")).Join(",")
Supports up to 100 images per request
Useful for visual analysis, OCR, image descriptions
Coda Webhook URL (Optional): URL for instant notification when processing completes
Format: Copy from automation webhook trigger
Faster than polling with "Check for Response"
See Coda webhook documentation for setup
Leave blank to use polling method
Coda API Token (Optional): Token for webhook authentication
Only needed if using webhook URL
Create at:
https://coda.io/account?openApiSettings=true
⁠
⁠
Keep secure - visible to anyone with doc access
Claude API Key Override (Optional): Override authentication key
Only use if Coda Pack auth is truncating your API key
Should start with sk-ant- and be ~108 characters
Most users won't need this
Click “Generate Response” - Wait a few minutes. Check the

And please send feedback to jondallasjr@gmail.com! Thank you very much 🙏
claudeAsync()
⁠
Logs
Timestamp
Function
Log
10/21/2025, 11:54 AM
Initiating prompt request
10/21/2025, 11:54 AM
Checking for response
10/21/2025, 11:55 AM
Checking for response: Attempt #2
10/21/2025, 11:55 AM
Checking for response
10/21/2025, 11:56 AM
Checking for response: Attempt #3
10/21/2025, 11:56 AM
Checking for response
10/21/2025, 11:57 AM
Checking for response: Attempt #4
10/21/2025, 11:57 AM
Checking for response, attempt #4
10/21/2025, 11:58 AM
Checking for response, attempt #5
10/21/2025, 11:59 AM
Checking for response, attempt #6
10/21/2025, 11:59 AM
Request retrieve
10/21/2025, 12:05 PM
Initiating prompt request
10/21/2025, 12:05 PM
Checking for response, attempt #1
10/21/2025, 12:06 PM
Checking for response, attempt #2
10/21/2025, 12:07 PM
Checking for response, attempt #3
10/21/2025, 12:08 PM
Checking for response, attempt #4
10/21/2025, 12:09 PM
Checking for response, attempt #5
10/21/2025, 12:10 PM
Checking for response, attempt #6
10/21/2025, 12:11 PM
completed
There are no rows in this table
⁠

Want to print your doc?
This is not the way.
Try clicking the ⋯ next to your doc name or using a keyboard shortcut (
CtrlP
) instead.