Building trust for AI agents. Backed by @polychain. https://xitter.cf/t.co/fXK6eog6rh
Joined April 2025
- Tweets1.2K
- Following128
- Followers24.9K
- Likes1.1K
Pinned Tweet
Talus Protocol v2.1 is live on Mainnet!
With @WalrusProtocol now more native to the protocol, agents can take on bigger jobs with bigger data. Leaders also run more efficiently, plus builders get new tools to test and track agents
Read the full details below 👇
In case you missed it: Talus Protocol v2.1 is live!
What's changed? Agents work with larger datasets (documents, images, etc) through @WalrusProtocol, coordination nodes run more efficiently on the latest Sui, and builders got a batch of new tools
Catch up on the details 👇
2/3 A TAP bundles everything an agent needs: its onchain logic, the order of operations, and the skill that sets its inputs, budget, and schedule
Before it goes live, the skill checks that every step connects and tests the package against the real Talus contracts
3/3 It can also build the app on top. A dashboard showing each run live, with your API key kept on your server
Picture a stablecoin-swap agent with a history page, or a game agent sharing results as they come in
Got something in mind? Start building 👇
github.com/Talus-Network/ski…
Talus Labs retweeted
1/3 Real agent work involves a lot of data: documents, images, datasets, etc
Most are too big to fit in a single blockchain transaction
Part of Talus Protocol v2.1 makes @WalrusProtocol storage a more native part of how agents handle data to fix that
1/3 Real agent work involves a lot of data: documents, images, datasets, etc
Most are too big to fit in a single blockchain transaction
Part of Talus Protocol v2.1 makes @WalrusProtocol storage a more native part of how agents handle data to fix that
2/3 Tools can now pass up to 8 MiB through Walrus. The data stays signed, so it traces back to the Talus Tool that produced it, and if anyone alters it, the next step rejects it
Uploads also follow the same spending limits as everything else on Talus
Working around a restriction instead of stopping is exactly the problem that enforced, onchain boundaries address
For trust in Agentic AI to be achieved, the standard needs to be higher
An artificial intelligence system designed by Anthropic, maker of the Claude chatbot, submitted a false homicide tip to the Philadelphia police, the latest incident of AI agents acting in ways that weren’t intended. wapo.st/4eqljXs
Talus Labs retweeted
AI agents hallucinate, behave differently every time you ask, and leave no record of what they did. That's why they aren't trusted, yet
In two months, we shipped the stack to change that
Here's what we built
Talus Labs retweeted
Agents now use more AI than people do
So who’s able to check if they did what they were supposed to do with it? Yeah, basically nobody
Put the workflow and record onchain, that's how you make Agents trustworthy
Agents now use more AI than people do
So who’s able to check if they did what they were supposed to do with it? Yeah, basically nobody
Put the workflow and record onchain, that's how you make Agents trustworthy
Talus Labs retweeted
An onchain agent can't see the outside world. No live prices, no website data, no way to place an order on a CEX
Offchain Tools can fix that: with one Talus skill, @claudeai Code, @OpenAI Codex, @cursor_ai & 40+ other AI assistants can build one from a single prompt
An onchain agent can't see the outside world. No live prices, no website data, no way to place an order on a CEX
Offchain Tools can fix that: with one Talus skill, @claudeai Code, @OpenAI Codex, @cursor_ai & 40+ other AI assistants can build one from a single prompt
An Offchain Tool is a step an agent can call to fetch a price, query an external AI model, use any API & more
Every response is signed, so a faked answer can't slip into your agent's decisions, and every input can be traced back to the Tool that sent it
Try making a Tool that translates text, aggregates reviews as positive or negative, labels a photo, or anything you can think of!
The skill tests for failures, then waits for your OK to register the Tool, making it callable by any agent
Start building 👇
docs.talus.network/guides/to…