>What Is Pica? Pica is an eating disorder in which a person eats things not usually considered food. Young kids often put non-food items (like grass or toys) in their mouths because they're curious about the world around them. But kids with pica (PIE-kuh) go beyond that.
Pica officially stands for "Platform for Intelligent, Collaborative Agents," reflecting our mission to enable seamless collaboration between AI systems. When we chose the name, we loved that picaos.com was available. While we’re aware of the condition “pica,” we like to think our platform has an insatiable appetite for APIs, integrations, and data—thankfully, no chalk or clay involved! :D
Oops, thanks for pointing that out! Pica actually stands for Platform for Intelligent, Collaborative Agents—but as with many words, it seems to have multiple meanings worldwide! We'll take this as a reminder to keep our humor handy while we chase big AI dreams. Appreciate the feedback!
haha, you got me! I totally panicked in the moment and asked AI for a quick reply, then had a good laugh about it. Naming across languages can be tricky. Thanks for keeping me on my toes ;P
I might be biased but I like the name. To add to the commentary with a positive connotation, Pica pica is the Latin name for the Eurasian magpie, perhaps the smartest bird (and animal?) of all :-)
Like you, I am biased, but I too love the name. And you are right Pica pica is definitely on the list of smartest animals. Thanks for your support and kind words :)
AI agents excel in tasks like automating workflows, handling complex decision trees, and managing multi-step processes across APIs. For example, an agent can monitor a sales pipeline, send follow-ups, update CRMs, or manage logistics autonomously. See our demo here: https://github.com/picahq/onetool-demo.
this is cool - looks like lots of the connectors are not available though? I work on https://mcp.run, which has some overlap to pica, where its really easy to ship your own connectors and run them directly in your LLM/Agent, or compose them into a Task.
would love to connect and see if there are ways to share :)
there are problems in ml-infra where rust could have a good impact. appears to me that there you are merging 2 popular bandwagons - agentic and rust - perhaps you could solve specific infra problems using rust which are inefficiently solved with python.
Agree. We use Rust to build highly performant systems at the lowest level, removing any language-related friction. Our system enables better management and auditing of agent interactions, leading to more effective activations and greater overall impact.
This is really cool! I can see the list of your supported Connectors on the site. How could I create custom integrations in Pica that would be accessible via the AI SDK?
Great question! We’ve developed a CLI specifically for creating and managing Connectors in Pica. It allows you to spin up Pica and seamlessly build integrations tailored to your needs. You can get started here: https://www.npmjs.com/package/@picahq/cli
Thank you for sharing your perspective! We chose Pica because it stands for "Platform for Intelligent, Collaborative Agents" and that really captures our vision.
We recognize naming can be tricky, words carry different connotations across cultures and languages. Still, we’re excited about what Pica will bring to the table, and we hope you’ll judge us more by our product than our name. Thanks again for voicing your thoughts.
Pica's OneTool integrates with the Vercel AI SDK and LangChain, simplifying API connections for your LLM. It enhances AI capabilities without expanding the context window, reducing tool-related hallucinations. For example, to work with 40 tools, you'd typically need 4 agents and a router agent. With Pica, a single agent can access all 40 tools simply by adding OneTool to your LLM.
It works very similarly to RAG but for tools. The SDK loads the necessary configs and schemas for each tool and uses that to guide the AI on how to use the tool effectively based on the latest message.
What does "in Rust" add to the stack that makes it special? From the title, I expected to see something in Rust on the home page, not JavaScript
How is this different from other AI agent infrastructure solutions out there? What particular pain point are you solving for your customers which won't be solved unless they switch?
Also OT but having to quickly hit "back" twice to get back from the website is mildly infuriating
The core system is built in Rust for performance, safety, and scalability. Integrations and tools are handled via JSON configs and JS mapping, while the interaction happens through our JS SDKs (Python and more languages coming soon).
We solve three key problems:
1) Simplified API/tool access without blowing up context windows.
2) Full auditability of AI actions for trust and oversight.
3) Seamless human-aligned guardrails for safe autonomy.
For example, with Pica's OneTool, a single agent can manage 40 tools without needing multiple agents and routers, reducing complexity and hallucinations.
The current connector allows your LLM to pass any query to your PostgreSQL database. For example, If you have read-only access to your database, you can connect it to Pica and ask the AI agent to list all your tables—it can do that seamlessly. Afterward, you can ask the agent to perform any query, and it will write the query, pass it to Pica via OneTool, and return the results to you.
As a Python developer, I often see a serious lack of even basic software engineering practices affecting code maintainability in commercial software. Is this less of an issue in Rust?
Unfortunately, no—it's not inherently less of an issue in Rust. It really comes down to the team and their passion for writing maintainable, high-quality code. The language helps, but good practices ultimately rely on the people behind the code.
Haha! We're actively enhancing our SDK to enable granular permissions for Connectors, giving you complete control over exactly what your Agent can access!
There seems to be a common medical definition for the word Pica
From: https://kidshealth.org/en/parents/pica.html#:~:text=What%20I....
>What Is Pica? Pica is an eating disorder in which a person eats things not usually considered food. Young kids often put non-food items (like grass or toys) in their mouths because they're curious about the world around them. But kids with pica (PIE-kuh) go beyond that.
FYI
Pica officially stands for "Platform for Intelligent, Collaborative Agents," reflecting our mission to enable seamless collaboration between AI systems. When we chose the name, we loved that picaos.com was available. While we’re aware of the condition “pica,” we like to think our platform has an insatiable appetite for APIs, integrations, and data—thankfully, no chalk or clay involved! :D
In Brazil, pica is a slang for penis.
Oops, thanks for pointing that out! Pica actually stands for Platform for Intelligent, Collaborative Agents—but as with many words, it seems to have multiple meanings worldwide! We'll take this as a reminder to keep our humor handy while we chase big AI dreams. Appreciate the feedback!
Also, the “ao” (actually ão) suffix in brazilian portuguese means “a huge version of” - so the URL picaos means “a lot of large penises”
Not that this matters at all, the platform seems nice and we brazilians love this kind of humor
Are you using AI to respond to these comments? Yikes.
haha, you got me! I totally panicked in the moment and asked AI for a quick reply, then had a good laugh about it. Naming across languages can be tricky. Thanks for keeping me on my toes ;P
Double down, baby!
And in most slavic languages its slang for the opposite organ
I might be biased but I like the name. To add to the commentary with a positive connotation, Pica pica is the Latin name for the Eurasian magpie, perhaps the smartest bird (and animal?) of all :-)
Congratulations on the launch, best of luck.
Like you, I am biased, but I too love the name. And you are right Pica pica is definitely on the list of smartest animals. Thanks for your support and kind words :)
What are the hero use cases for AI agents? Reading the website, I don't see a good demonstration of what "AI agents" are and what they're good for.
AI agents excel in tasks like automating workflows, handling complex decision trees, and managing multi-step processes across APIs. For example, an agent can monitor a sales pipeline, send follow-ups, update CRMs, or manage logistics autonomously. See our demo here: https://github.com/picahq/onetool-demo.
This particular Oxide and Friends podcast includes a fantastic discussion of AI agents, the problems with their definitions, and at least one promising use case: https://oxide-and-friends.transistor.fm/episodes/predictions...
this is cool - looks like lots of the connectors are not available though? I work on https://mcp.run, which has some overlap to pica, where its really easy to ship your own connectors and run them directly in your LLM/Agent, or compose them into a Task.
would love to connect and see if there are ways to share :)
I would love to connect! We currently have about 40 connectors ready to use, and both our team and the community are actively adding more as we go.
btw, we will be launching an MCP support in the coming weeks.
I’m excited to learn more about how we could collaborate to benefit the community :)
Here’s the link to explore our connectors: https://docs.picaos.com/api-reference/connectors Feel free to grab a time on my calendar here: https://calendly.com/moekatib/founder-coffee
awesome -- will grab some time!
Is the OneTool source not available?
Also, do you guys do something similar to: https://github.com/frankfralick/dripgrep ?
(for the curious, the README for this repo is 100% worth taking a look at!)
The source code for OneTool is available here: https://github.com/picahq/ai
I'll definitely check it out—thanks for sharing!
there are problems in ml-infra where rust could have a good impact. appears to me that there you are merging 2 popular bandwagons - agentic and rust - perhaps you could solve specific infra problems using rust which are inefficiently solved with python.
Agree. We use Rust to build highly performant systems at the lowest level, removing any language-related friction. Our system enables better management and auditing of agent interactions, leading to more effective activations and greater overall impact.
Back button is nonfunctional (because of the redirect to /home)
The fix has been deployed
This is really cool! I can see the list of your supported Connectors on the site. How could I create custom integrations in Pica that would be accessible via the AI SDK?
Great question! We’ve developed a CLI specifically for creating and managing Connectors in Pica. It allows you to spin up Pica and seamlessly build integrations tailored to your needs. You can get started here: https://www.npmjs.com/package/@picahq/cli
Whatever the utility may be, the naming is really unfortunate and offputting IMO.
Thank you for sharing your perspective! We chose Pica because it stands for "Platform for Intelligent, Collaborative Agents" and that really captures our vision.
We recognize naming can be tricky, words carry different connotations across cultures and languages. Still, we’re excited about what Pica will bring to the table, and we hope you’ll judge us more by our product than our name. Thanks again for voicing your thoughts.
Do you really believe letting AI write your replies come off as professional sounding?
Can you explain what value this provides over just the Vercel AI SDK? I'm not super clear on what it is based on the homepage.
Pica's OneTool integrates with the Vercel AI SDK and LangChain, simplifying API connections for your LLM. It enhances AI capabilities without expanding the context window, reducing tool-related hallucinations. For example, to work with 40 tools, you'd typically need 4 agents and a router agent. With Pica, a single agent can access all 40 tools simply by adding OneTool to your LLM.
See how it works with Pica here https://github.com/picahq/onetool-demo
I'm guessing it's basically doing RAG over the full toolset to surface only the most relevant tools for any given message?
It works very similarly to RAG but for tools. The SDK loads the necessary configs and schemas for each tool and uses that to guide the AI on how to use the tool effectively based on the latest message.
Congrats on the launch and thanks for posting!
What does "in Rust" add to the stack that makes it special? From the title, I expected to see something in Rust on the home page, not JavaScript
How is this different from other AI agent infrastructure solutions out there? What particular pain point are you solving for your customers which won't be solved unless they switch?
Also OT but having to quickly hit "back" twice to get back from the website is mildly infuriating
Thank you for the thoughtful questions :)
The core system is built in Rust for performance, safety, and scalability. Integrations and tools are handled via JSON configs and JS mapping, while the interaction happens through our JS SDKs (Python and more languages coming soon).
We solve three key problems:
1) Simplified API/tool access without blowing up context windows. 2) Full auditability of AI actions for trust and oversight. 3) Seamless human-aligned guardrails for safe autonomy.
For example, with Pica's OneTool, a single agent can manage 40 tools without needing multiple agents and routers, reducing complexity and hallucinations.
Feedback on the site is noted—thank you!
Can you please provide a really basic example of how it can be used with PostgreSQL?
The current connector allows your LLM to pass any query to your PostgreSQL database. For example, If you have read-only access to your database, you can connect it to Pica and ask the AI agent to list all your tables—it can do that seamlessly. Afterward, you can ask the agent to perform any query, and it will write the query, pass it to Pica via OneTool, and return the results to you.
We’ll also be adding many use cases to our community hub in the coming days, so stay tuned! https://www.picaos.com/community/use-cases
Thanks! Will give it a try.
lá ele ( ͡° ͜ʖ ͡°)
As a Python developer, I often see a serious lack of even basic software engineering practices affecting code maintainability in commercial software. Is this less of an issue in Rust?
Unfortunately, no—it's not inherently less of an issue in Rust. It really comes down to the team and their passion for writing maintainable, high-quality code. The language helps, but good practices ultimately rely on the people behind the code.
Super cool! Also somewhat terrifying—the comment text on the landing page definitely raises an eyebrow:
Haha! We're actively enhancing our SDK to enable granular permissions for Connectors, giving you complete control over exactly what your Agent can access!