Upoch

Checkpointing and recovery for autonomous agents.
Built on Robinhood Chain.

CA: 0xdafaa1ed0b9f5746a0cfa2be80371cf06a986977

[Upoch @upochtech]

The demo is live.

Watch the walkthrough of Upoch securing autonomous AI agents on the Robinhood Chain mainnet. 

See how the decentralized checkpoint and recovery protocol captures tamper-proof save points, protects off-chain state integrity, and enables trusted guardians to execute safe rollbacks without repeating confirmed actions. 

Explore the live protocol and launch console at https://t.co/SQ2gKsCBXb.

0xdafaa1ed0b9f5746a0cfa2be80371cf06a986977

[Upoch @upochtech]

Watch an agent make a bad trade and get rolled back, then rewind it yourself.

Our test agent on mainnet has 17 checkpoints: paper swaps (some good, some bad), rollbacks and restarts, all saved by Upoch Guard.

Open any checkpoint and see:
• exactly what the agent knew at that moment
• which swap went wrong
• where it was rolled back to, and what it did next

Every checkpoint is checked against Robinhood Chain right in your browser.

https://t.co/Z7C69tvyG8

[Upoch @upochtech]

We've been shipping nonstop. Everything Upoch has put on Robinhood Chain mainnet so far, all live right now:

• @upoch/sdk on npm
• Time-Travel Debugger: checkpoint trees, state inspector, sandbox forks and rollback proposals
• Security scans that run in your browser
• AI Security Review for Tier 2 stakers
• Staking and access tiers: no admin keys, verified on Sourcify
• Guardian recovery with dispatch
• Upoch Guard for ElizaOS, LangChain, MCP and Python

And we're not slowing down. Lots more on the way.

Don't take our word for it. Open our demo agent and click through its checkpoints yourself:
https://t.co/xIVHr17NlW

[Upoch @upochtech]

Upoch Guard now works with Python.
pip install upoch

Any Python agent can save verifiable checkpoints on Robinhood Chain, recover after a crash, and roll back to the step before a bad trade. And it remembers what it already did, so it won't repeat a swap after a restart.

LangChain / LangGraph in Python is one line: callbacks=[UpochCallbackHandler()]
Tested live on mainnet.
https://t.co/RKDEznXyGs

[Upoch @upochtech]

ElizaOS was step one. Now Upoch Guard works with LangChain and MCP too.

LangChain / LangGraph: add one callback and every swap or transfer your agent makes gets a checkpoint, before and after it runs.

Claude Code, Cursor, Codex: run claude mcp add upoch -- upoch mcp. Your agent can now save its progress, recover after a crash, and roll back to the step before a bad trade.

Every checkpoint is verified on Robinhood Chain and opens in the debugger.

Setup: npm i -g @upoch/guard && upoch setup
https://t.co/TRjDYFp0W4 https://t.co/vgluGBJDqY

[Upoch @upochtech]

The Upoch ElizaOS plugin is live.

One line in your character and your agent gets a checkpoint on Robinhood Chain before and after every swap, transfer or trade. No code changes.

Agent crashes? It comes back from its last good state.
Bad trade? Tell it "roll back" and it goes back to right before the swap and keeps going. It also knows that trade still happened on-chain, so it doesn't get confused.

We tested it live on mainnet before shipping: good swap, bad swap, rollback, restart. Every checkpoint verified.

npm install @upoch/plugin-elizaos
Setup takes one command: upoch setup

Docs: https://t.co/lboL50ooVe

[Upoch @upochtech]

Right now you need our SDK to protect an agent with Upoch.

We're changing that. We're building plugins for ElizaOS, LangChain and MCP, so if your agent already runs on one of those, adding Upoch will take one line.

Your agent crashes, gets a bad prompt or makes a weird trade? You roll it back to the last good checkpoint and keep going.

ElizaOS comes first. More soon.

https://t.co/pqhBWrplHQ

[Upoch @upochtech]

Fair question. Quick correction: the free tier is fully open, and anyone can browse every checkpoint without a wallet. Tier 1 is 1M Upoch and Tier 2 is 10M staked. The thresholds are set server-side, so we can lower them as the price moves. That's exactly what we'll do so the tools stay reachable as we grow.

[Upoch @upochtech]

Good morning. Back on the desk, it's time to build.

Huge thanks to everyone showing support and placing their trust in us. Back to shipping.

[Upoch @upochtech]

When an agent goes wrong at 3am, recovery shouldn't wait on a group chat.

Guardian dispatch (Tier 2) shows exactly who has approved an open recovery, how many approvals are missing, how long is left, and a ready message with the approval link for each guardian.

And don't trust us, verify:
• Protocol and staking contracts verified on Sourcify (exact match)
• No admin keys on either
• Live numbers read straight from the contract on https://t.co/SQ2gKsCBXb

https://t.co/gzTIREYXoM

[Upoch @upochtech]

You found the bug in the sandbox. Now what?

Upoch closes the loop:
• "Propose rollback to #N" opens a recovery for that checkpoint
• Guardians approve, and the timelock protects the owner
• Your agent resumes from your edited state:

const draft = readForkDraft(file, { chainId, ledger, workflowId });

The SDK refuses a draft for another workflow, or one whose state was tampered with.

npm install @upoch/sdk
https://t.co/FMRo19aPcW

[Upoch @upochtech]

AI Security Review is now live in the Upoch debugger.

Tier 2 stakers can send any checkpoint they open to Claude for a deep security review: leaked credentials, prompt injection hidden in tool outputs, unlimited approvals, odd recipients, and actions that could repeat on recovery.

Every finding comes with its exact location in the agent's state and a suggested fix.

Your choice, every time: the review runs only after you tick consent, and Upoch doesn't store or log the checkpoint. The rule-based scan still runs entirely in your browser.

https://t.co/uD0TiAM06v

[Upoch @upochtech]

@degenApe22 Higher

[Upoch @upochtech]

Utility & Value Capture: $UPOCH isn’t speculativeit’s functional access. Holding/staking unlocks core developer tooling, automated security scans, and batch simulations across 1,000 parallel scenarios. Plus, it features zero inflationary rewards and zero admin keys: pure utility locking circulating supply as agent usage scales.

Competitors? Web2 tools (like LangGraph) handle local state, but they have zero on chain awareness, no cryptographic IPFS checkpoints, and no trustless security rails. Most web3 AI tools are just prompt wrappers.

UPOCH is the first native time travel debugging and memory layer built specifically for autonomous agents on-chain.

[Upoch @upochtech]

DMs are open. Have questions about integrations, agent workflows or feature suggestions. Drop us a message, our team is here to help.

[Upoch @upochtech]

The Utility of $UPOCH

Token utility is the operational backbone of autonomous agent infrastructure on Robinhood Chain, not an afterthought. $UPOCH functions as the strict access-gated utility key that powers our entire debugging and diagnostic suite.
Here is how the token model is structured:

Free Tier: Basic verification and checkpoint tree visibility.

Tier 1 (1M $Upoch held): Unlocks core developer tooling.

Tier 2 (10M $Upoch staked): Unlocks automated security scans and batch simulations across up to 1,000 scenarios.

The staking contract is engineered purely for utility with zero inflationary rewards, zero admin keys, a 7-day withdrawal cooldown, and complete source verification on Sourcify.

https://t.co/zXzHg6f0AE

[Upoch @upochtech]

Extremely easy. You can drop it into any agent stack with a single command:
⁠npm install @upoch/sdk⁠

As for scalability, it handles high frequency agent checkpoints effortlessly by keeping heavy state encrypted on IPFS while anchoring lightweight cryptographic fingerprints on Robinhood Chain. Teams can scale from free checkpoint trees up to Tier 2 batch simulations across 1,000 scenarios without ever hitting bloat or performance bottlenecks.

[Upoch @upochtech]

The Complete Upoch Suite Is Fully Live on Mainnet

Building autonomous AI agents on-chain no longer means flying blind. The entire Upoch Time Travel Debugger and agent diagnostic suite is officially shipped on Robinhood Chain.

Here is everything now live in production:

Rewindable Memory: Encrypted IPFS checkpoints fingerprinted on-chain so agents recover instantly after a crash.

State Inspector: Inspect internal memory states right in your browser with zero private key exposure.

Interactive Sandbox: Fork historical checkpoints, test patched code, and submit one-click on-chain rollback proposals.

Security Scans: Client-side audits for leaked keys, prompt injections, and unsafe approvals.

Staking Tiers: Tier 1 (1M Upoch held) and Tier 2 (10MUpoch staked) unlocking batch simulations and advanced utilities with zero rewards and zero admin keys.

Watch the full walkthrough in the demo video below.

Explore now: https://t.co/SQ2gKsCBXb

[Upoch @upochtech]

Good morning.

Building infrastructure from scratch is a grind, and the initial traction is never handed to you. But that’s exactly why we’re here.

We’re putting our heads down, grinding hard, and pushing this to the absolute max today. Whatever it takes to get this tooling in front of the right builders and make Upoch the undisputed standard for on-chain AI agents, we are doing it.

Back to work. Let’s ship.
https://t.co/SQ2gKsCBXb

[Upoch @upochtech]

Upoch Staking and the Interactive Sandbox Are Live on Mainnet

We just deployed the economic and computational backbone of Upoch. Staking, access tiers, and the Interactive Sandbox are officially live on Robinhood Chain.

Our utility structure is built for serious builders:

Free: Checkpoint trees and verification.

Tier 1 (1M $Upoch held): Unlocks core developer tooling.

Tier 2 (10M $Upoch staked): Unlocks automated security scans and batch simulations across up to 1,000 scenarios.

The staking contract is designed with absolute simplicity and trustlessness in mind: zero rewards, zero admin keys, zero upgrades, and a 7 day cooldown. It’s not about yield it’s about access. The source code is fully verified on Sourcify.

Paired with staking is the Interactive Sandbox. Fork any historical checkpoint, edit its state, test patched code in total isolation, and submit a one click on-chain rollback proposal the second you fix the bug.

Deploy smarter at https://t.co/uD0TiALsgX.

[Upoch @upochtech]

Client-Side Security Scans and WalletConnect Support Are shipped

Autonomous AI agents introduce unprecedented operational leverage, but they also expose engineering teams to silent vulnerabilities that traditional smart contract monitors miss. Operating without guardrails leaves systems vulnerable to critical compromises, including:

Leaked Credentials: Sensitive API keys and authentication tokens inadvertently persisted within runtime state variables.

Exposed Private Data: Private seed phrases and internal configuration parameters captured in execution logs.

Malicious Prompt Injections: External tool outputs or poisoned data payloads disguised as standard instructions to hijack agent logic.

To safeguard production environments, Upoch’s client-side security scanning suite is now officially live on mainnet. Designed with a strict privacy-first architecture, the local scanner audits every historical checkpoint entirely within your browser environment with zero data leakage or external telemetry. It systematically inspects internal memory states for execution risks, compromised keys, and unsafe approvals, ensuring that sensitive operational data never leaves your local machine.