This thing is amazing. THough pricing is little high and only values annual users, but hopefully its more flexible in pricing. But damn it works like magic for short work flows. Though complicated ones are still somewhat lacking but great one!.
This is fascinating! 🚀 The ability to create agents for data processing without API keys is impressive. How do you handle updates or changes in the APIs of integrated services like Snowflake or Postgres? 📊 Eager to see how it performs!
@jaredl Updates or changes to APIs do not impact deployed agents (unless, of course, the update is a breaking change for prior APIs). We do offer a clean upgrade path to adopt updates or changes if you want.
Report
Ohhh! String.com is a total lifesaver! how come i didn't stumble on this earlier? so hyped rn!
Tried tools like LangChain before—they work, but String.com? way simpler. no - code builders feel clunky next to this. string just gets the agent built in seconds, lol. it’s like langchain on espresso, way more efficient.
gonna use it to build an agent that monitors our team’s notion docs and auto - creates jira tickets. no more manual tracking headaches! and later, test web scraping agents for market research. finally, a tool that turns “what if” prompts into real, working agents without the setup mess.
congrats to the string.com team! this is gonna make so many devs and product folks skip the usual agent - building pain. let’s prompt - build some magic.
@sumitgoel we're definitely thinking about how String can work with your own AI agent. Please book some time on my calendar if you want to discuss https://calendly.com/pipedream/30-minute-meeting -- I'd love to learn more about your use case. We also launched https://pipedream.com/connect so your agents can use 2700+ apps and 10k+ tools on behalf of your users (they don't need their own own Pipedream accounts).
This is very cool! I'm gonna try building an agent that researches free trial signups for my product based on some specific parameters and send me slack notifications for the signups I should be focused on to convert them to paid customers!
Like many folks I use Trello to simplify managing my tasks. I recently tried out String.com to build an AI agent designed to automatically analyze unlabeled Trello cards and apply relevant labels. This is especially helpful as I use email-to-board to automatically create cards for some requests. This workflow would be incredibly helpful, imagine a perfectly organized Trello board without any manual labeling effort! My goal was to have it analyze existing unlabeled cards and apply appropriate labels based on what was already on the board.
String.com's marketing promises a revolutionary approach: "What if an AI agent could build AI agents? Just type a prompt — String builds the agent for you. Way easier than no-code builders. Handles way more use cases with real, working code." They boast integration with "thousands of apps" and "AI integration with OpenAI... batteries included, no API keys required. No setup hell. No keys to manage. Just click to auth, test, and deploy."
The initial setup seemed to align with this promise: I gave it this prompt “Take unlabeled cards on a Trello board analyze them with AI and apply the most applicable labels.” String.com then created a plan, I approved the plan, the agent was created, and the Trello component was quickly configured.
Unfortunately, despite the marketing claims of "no setup hell" and "batteries included," the AI seemed to encounter a lot of errors, though it was able to move on from those errors on its own. This resulted in a lot of repeated, very inefficient requests, using more than 2 million tokens without successfully producing anything. I’m out of tokens on my paid plan, so I can’t finish this up for another 26 days.
Here's what I liked:
Great Idea: The concept of using AI to build an AI powered tools is great.
Intuitive Setup: Agent configuration and Trello connection were straightforward, fulfilling "just type a prompt" and "easier than no-code builders" claims.
Self-repair: The agent diagnosed and recovered from errors autonomously, indicating strong underlying intelligence.
However, I encountered significant issues that highlight areas for improvement:
Excessive AI token consumption was the critical issue. Despite only 8 unlabeled Trello cards, the agent used over 2 million AI tokens, depleting my supply and halting the process. This contradicts "batteries included" and "no setup hell" claims, as such high consumption for a minimal task is highly inefficient and prohibitively costly for production.
Early Challenges and Unforeseen Expenses: Initial friction points included the retrieval of "0 cards" and a "Test failed" step. The agent's multiple unsuccessful attempts undeniably consumed tokens. These repeated efforts contributed to overall inefficiency and an unexpected depletion of my paid plan's allowance.
Comparison: I attempted to create the same project with Gemini-CLI, for it I used the prompt "Create a python script which retrieves all unlabeled cards from a trello board analyze them with AI and apply the most appropriate labels." The first 3 turns were spent on setting up the python environment, which I could have done manually with 'uv init' but I wanted to see what the AI would do. I then spent another 3 turns refining and polishing the script and now have a reusable package. I included the token usage stats below from /stats in Gemini-CLI. This seems like reasonable token usage for the amount of work that was done.
String.com's new product is super innovative, and I'm really optimistic about its potential! Right now, there are some kinks to work out for it to be truly viable. But after watching what they’ve built with Pipedream over the past few years I have a good feeling the team at String.com will tackle these issues soon.
@jnhaisley thanks for the detailed write up! This is exactly the type of feedback we're looking for during the early-preview. We're continuously tuning which cases String solves efficiently and which need more work. That said, we're super excited about what String can already do today and we're even more excited about the future possibilities.
I'd like to offer you more tokens so you can keep using the product and give us more feedback. I believe we've connected before -- I'd love to chat more if you're open to it. You can book time on my calendar here https://calendly.com/pipedream/30-minute-meeting. Or just ping me in our public Slack or reach out to support@pipedream.com and we'll load up more AI tokens for you!
Report
@pravinsavkar Thanks for reaching out! I sent you a message on slack, and I'm super excited about the possibilities, and I know after watching the evolution of pipedream what your team is capable of producing.
Report
This was so promising that I was ready to get our agency to switch over and to onboard our clients - IF it worked. Unfortunately nearly 4million tokens in and after very basic mistakes and lack of complying with our instructions I'm now being prompted to pay for something I am not sure will work. I would love for the tool to clarify things before jumping straight into building wherever it may be confused. Also connecting to the dependencies and checking the architecture of (for example our CRM) would be a big boost in saving tokens/credits.
Would love to be a paid user but not sure if it works, why should I take a chance on this?
Also, I'd be happy to give you more AI tokens to see if that helps get you to value. Feel free to ping me in our public Slack or reach out to our support team and point them to this thread.
What if an AI agent could build AI agents?
Introducing String.com by Pipedream.
Just type a prompt — String builds the agent for you.
Way easier than no-code builders.
Handles way more use cases with real, working code.
Example? Try, "Build me an agent that watches a GitHub repo and creates matching Linear issues."
Done in one shot.
But String.com isn’t just for simple stuff — chat to generate full-featured agents using custom code and tools for thousands of apps:
Messaging with Slack or Discord
Internal tools like Airtable, Notion, and Google Sheets
Data processing with Snowflake, Postgres, or BigQuery
Web automation like scraping, browsing, and monitoring
AI integration with OpenAI, Anthropic, and others — batteries included, no API keys required
No setup hell. No keys to manage. Just click to auth, test, and deploy.
Try it free. See what one prompt can do.
String.com
This is fascinating! 🚀 The ability to create agents for data processing without API keys is impressive. How do you handle updates or changes in the APIs of integrated services like Snowflake or Postgres? 📊 Eager to see how it performs!
@jaredl Updates or changes to APIs do not impact deployed agents (unless, of course, the update is a breaking change for prior APIs). We do offer a clean upgrade path to adopt updates or changes if you want.
@sumitgoel we're definitely thinking about how String can work with your own AI agent. Please book some time on my calendar if you want to discuss https://calendly.com/pipedream/30-minute-meeting -- I'd love to learn more about your use case. We also launched https://pipedream.com/connect so your agents can use 2700+ apps and 10k+ tools on behalf of your users (they don't need their own own Pipedream accounts).
This is very cool! I'm gonna try building an agent that researches free trial signups for my product based on some specific parameters and send me slack notifications for the signups I should be focused on to convert them to paid customers!
@neel_balar Please let us know how it goes!
Promising Concept, But Major Efficiency Hurdles
Like many folks I use Trello to simplify managing my tasks. I recently tried out String.com to build an AI agent designed to automatically analyze unlabeled Trello cards and apply relevant labels. This is especially helpful as I use email-to-board to automatically create cards for some requests. This workflow would be incredibly helpful, imagine a perfectly organized Trello board without any manual labeling effort! My goal was to have it analyze existing unlabeled cards and apply appropriate labels based on what was already on the board.
String.com's marketing promises a revolutionary approach: "What if an AI agent could build AI agents? Just type a prompt — String builds the agent for you. Way easier than no-code builders. Handles way more use cases with real, working code." They boast integration with "thousands of apps" and "AI integration with OpenAI... batteries included, no API keys required. No setup hell. No keys to manage. Just click to auth, test, and deploy."
The initial setup seemed to align with this promise: I gave it this prompt “Take unlabeled cards on a Trello board analyze them with AI and apply the most applicable labels.” String.com then created a plan, I approved the plan, the agent was created, and the Trello component was quickly configured.
Unfortunately, despite the marketing claims of "no setup hell" and "batteries included," the AI seemed to encounter a lot of errors, though it was able to move on from those errors on its own. This resulted in a lot of repeated, very inefficient requests, using more than 2 million tokens without successfully producing anything. I’m out of tokens on my paid plan, so I can’t finish this up for another 26 days.
Here's what I liked:
Great Idea: The concept of using AI to build an AI powered tools is great.
Intuitive Setup: Agent configuration and Trello connection were straightforward, fulfilling "just type a prompt" and "easier than no-code builders" claims.
Self-repair: The agent diagnosed and recovered from errors autonomously, indicating strong underlying intelligence.
However, I encountered significant issues that highlight areas for improvement:
Excessive AI token consumption was the critical issue. Despite only 8 unlabeled Trello cards, the agent used over 2 million AI tokens, depleting my supply and halting the process. This contradicts "batteries included" and "no setup hell" claims, as such high consumption for a minimal task is highly inefficient and prohibitively costly for production.
Early Challenges and Unforeseen Expenses: Initial friction points included the retrieval of "0 cards" and a "Test failed" step. The agent's multiple unsuccessful attempts undeniably consumed tokens. These repeated efforts contributed to overall inefficiency and an unexpected depletion of my paid plan's allowance.
Comparison:
I attempted to create the same project with Gemini-CLI, for it I used the prompt "Create a python script which retrieves all unlabeled cards from a trello board analyze them with AI and apply the most appropriate labels." The first 3 turns were spent on setting up the python environment, which I could have done manually with 'uv init' but I wanted to see what the AI would do. I then spent another 3 turns refining and polishing the script and now have a reusable package. I included the token usage stats below from /stats in Gemini-CLI. This seems like reasonable token usage for the amount of work that was done.
Stats Cumulative (6 Turns)
Input Tokens 464,165
Output Tokens 5,574
Thoughts Tokens 4,171
Cached Tokens 248,863 (52.5%)
Total Tokens 473,910
Overall:
String.com's new product is super innovative, and I'm really optimistic about its potential! Right now, there are some kinks to work out for it to be truly viable. But after watching what they’ve built with Pipedream over the past few years I have a good feeling the team at String.com will tackle these issues soon.
@jnhaisley thanks for the detailed write up! This is exactly the type of feedback we're looking for during the early-preview. We're continuously tuning which cases String solves efficiently and which need more work. That said, we're super excited about what String can already do today and we're even more excited about the future possibilities.
I'd like to offer you more tokens so you can keep using the product and give us more feedback. I believe we've connected before -- I'd love to chat more if you're open to it. You can book time on my calendar here https://calendly.com/pipedream/30-minute-meeting. Or just ping me in our public Slack or reach out to support@pipedream.com and we'll load up more AI tokens for you!
@pravinsavkar Thanks for reaching out! I sent you a message on slack, and I'm super excited about the possibilities, and I know after watching the evolution of pipedream what your team is capable of producing.
This was so promising that I was ready to get our agency to switch over and to onboard our clients - IF it worked. Unfortunately nearly 4million tokens in and after very basic mistakes and lack of complying with our instructions I'm now being prompted to pay for something I am not sure will work. I would love for the tool to clarify things before jumping straight into building wherever it may be confused. Also connecting to the dependencies and checking the architecture of (for example our CRM) would be a big boost in saving tokens/credits.
Would love to be a paid user but not sure if it works, why should I take a chance on this?
@jibril_ramadan1 thanks for the feedback! I'd love to connect to learn more about your use case and understand what's not working as expected. Can you book some time on my calendar here? https://calendly.com/pipedream/30-minute-meeting
Also, I'd be happy to give you more AI tokens to see if that helps get you to value. Feel free to ping me in our public Slack or reach out to our support team and point them to this thread.
@pravinsavkar Sure, Happy to help