COTI_Official

This week, a Claude Opus 4.8 audit found a four-year-old flaw in one of crypto’s most trusted privacy systems. Here is how and why COTI is built on a different foundation.
TL;DR
→ This week, Shielded Labs disclosed a critical bug in Zcash’s Orchard shielded pool, found with the help of Anthropic’s Claude Opus 4.8. → A single under-constrained element in a complex ZK circuit went undetected for four years, and privacy made it impossible to prove whether it was ever exploited. → COTI is built on a different foundation: Garbled Circuits, encrypted multi-party computation on battle-tested symmetric-key cryptography, with no soundness-critical proving circuit, no trusted setup, and quantum resistance by design. → COTI’s programmable privacy is verifiable by design. Selective disclosure lets supply and integrity be proven without exposing anyone’s data. → Private ERC20 supply stays governed and auditable.
A Wake-Up Call for Web3 Privacy
This week, the world of crypto privacy got a wake-up call.
Shielded Labs, the nonprofit that funds Zcash development, disclosed a critical soundness bug in Zcash’s Orchard shielded pool, one of the most respected zero-knowledge privacy systems in production.
The flaw was discovered using a custom auditing framework paired with Anthropic’s newly released Claude Opus 4.8. The flaw had been hiding in the proving circuit since Orchard launched in May 2022. For nearly four years, it sat undetected through repeated audits by some of the best cryptographers in the field.
The response was fast and coordinated. An emergency soft fork temporarily disabled Orchard transactions, and the hard fork corrected the circuit. That deserves real credit. Responsible disclosure, a patch within days, and full transparency are exactly how security is supposed to work.
Building trust within the privacy ecosystem is critical. A safe, reliable and secure privacy industry is key to uplift the entire ecosystem. Since blockchain privacy is a new breakthrough technology, it must continuously be hardened by the latest AI tools.
This is COTI’s design philosophy behind Garbled Circuits and the commitment to always ensuring industry-leading encryption standards.
Why This Matters Beyond Zcash
The story is bigger than any one network, and it lands on two hard truths every privacy protocol has to face.
Complex circuits are extraordinarily hard to get perfectly right. The flaw was an under-constrained element in the proving circuit: a check that was supposed to reject invalid inputs but contained a gap. Security researchers have long flagged under-constrained circuits as one of the most common failure classes in zero-knowledge audits, with the majority of ZK audit findings tracing back to the circuit layer.
Privacy can make verification impossible. Because Orchard is a shielded pool, Shielded Labs acknowledged there is no way to cryptographically determine whether the flaw was ever exploited. The same property that protects users also removes the trail that would let anyone prove the books are clean. You are left to take it on trust.
Claude Opus 4.8 audit found in about a day what years of expert human review missed. Every protocol will now be probed by the same class of tooling, by defenders and attackers alike, but simultaneously can use these tools to harden their infrastructure.
All privacy protocols must ensure their encryption and security standards are continuously improved and audited to ensure the highest-level of safety.
How COTI Is Built Different
COTI took a different cryptographic route from the ground up: Garbled Circuits, a protocol for secure multi-party computation (MPC), developed with Soda Labs.
Smart contracts compute directly on encrypted inputs, so the data is never exposed, not to an operator, not to the network, not to anyone who should not see it. That choice changes the security model in ways that matter right now.
MPC on encrypted inputs, not under-constrained ZK circuits
A zero-knowledge system proves a statement about hidden data using a complex circuit, and the soundness of that circuit is everything. One under-constrained constraint, as Orchard showed, can let invalid inputs pass.
COTI works differently. With Garbled Circuits, contracts compute on encrypted inputs through secure multi-party computation, and execution validity is checked by the COTI mainnet. There is no single soundness-critical proving circuit standing between you and the integrity of the whole pool.
Symmetric-key cryptography
COTI’s garbling scheme is built on the most studied, battle-tested primitives in cryptography. It applies a pseudo-random function such as AES to circuit wire labels, and on-chain ciphertexts are protected by a symmetric encryption scheme. These are the same families of primitives that secure the modern internet.
New privacy capabilities on top of trusted, well-understood building blocks, rather than a novel circuit that has to be proven flawless.
No trusted setup required
Some privacy systems depend on a one-time trusted setup ceremony. If the secret behind that ceremony is ever mishandled, the security of the whole system can be undermined. The Garbled Circuits approach requires no such ceremony. That removes an entire class of risk before a single transaction is processed.
Quantum-resistant by design
Because the foundation is symmetric-key cryptography, resistance to quantum attacks is a matter of key length, not a redesign. The best known quantum attacks on symmetric primitives use Grover’s algorithm, which roughly halves effective security, so doubling the key length restores it.
COTI’s Helium upgrade introduced native 256-bit arithmetic, so COTI’s confidential tokens already run at the key sizes that keep them secure against the best known quantum attacks. For the full breakdown, read COTI’s technical deep-dive on quantum computing.
And it is practical
The cryptography is strong, and it is usable. Confidential computation on COTI is up to 3,000x faster and 250x lighter than FHE-based approaches, light enough for any device including mobile, with no specialized hardware.
And COTI’s gcEVM is fully EVM-compatible, so developers write confidential contracts in standard Solidity, choosing what stays private and what goes public inside a single contract.
Verifiable by Design
The hardest part of this week’s story is not the bug. Bugs get patched. It is that no one can cryptographically prove whether the supply was ever inflated. That is the trap privacy systems can fall into: confidentiality that comes at the cost of verifiability.
COTI’s thesis has always been that transparency was never the real goal. The goal is the ability to verify information independently, which modern cryptography can deliver without exposing everything to the public.
With programmable privacy, COTI lets the right parties verify exactly what they need, supply, solvency, or compliance, through selective disclosure, while everyone else sees only ciphertext.
You can see it directly in Private ERC20s, COTI’s privacy-enhanced token standard:
Encrypted by default. Balances, transfer amounts, and allowances are encrypted. Each holder has their own key, and only the holder can decrypt their balance. Validators see ciphertext.
Private, but accountable, supply. totalSupply() returns zero on-chain, while a ciphertext version tracks the real supply internally. No leaking issuance to the outside world, and no blind spot either.
Governed issuance. Role-based access control, a dedicated minter role, and cap-aware issuance make supply governance explicit and auditable, with selective disclosure for compliance.
Proven foundations. Access control and reentrancy protection inherited from OpenZeppelin, encrypted arithmetic delegated to COTI’s MPC precompile, with full 256-bit precision.
This is privacy that does not force a choice between confidentiality and the ability to prove the truth. That is the difference between privacy you can verify and privacy you have to take on blind faith.
Audited End to End, and Leaning Into AI
Strong cryptography still has to be proven at the code level. No privacy technology is exempt from rigorous, ongoing review, which is exactly why COTI treats security as continuous.
Sayfer, a Web3-native cybersecurity consultancy that has secured billions in assets across 100+ clients including MetaMask, 1inch, Polkadot, and StarkWare, with zero client hacks to date, has audited COTI’s privacy stack line by line: Private ERC20, the Privacy Portal, and the COTI MetaMask Snap. Every finding was addressed and resolved before launch.
“After a review by the Sayfer team, we certify that all the security issues mentioned in this report have been addressed by the COTI team.” Sayfer audit report
Sayfer is building an AI-powered security layer for continuous, real-time auditing that moves past the point-in-time model, and COTI is a long-term partner in that work. When advanced AI red-teaming becomes standard across the industry, COTI intends to always utilize the latest technology
Live & In-Production Privacy.
Ethereum Co-founder Vitalik Buterin has publicly pointed to Garbled Circuits as the path to pure-cryptographic security guarantees for multi-party computation.
COTI’s GC mainnet launched in March 2025 and has processed 125 million+ on-chain transactions. COTI is bringing its GC-powered privacy protocol natively to Ethereum, and other leading L1s & L2s.
For major blockchains, DeFi, and dapps that need reliable, proven privacy that is audited and verifiable, COTI intends to be the industry-leader in trusted solutions.
Why COTI Is Unique
COTI is built so that trust in privacy does not depend on a promise:
Computes on encrypted inputs through MPC, with no single soundness-critical proving circuit.
Built on battle-tested symmetric-key cryptography.
Requires no trusted setup ceremony.
Quantum-resistant by design, with native 256-bit key sizes already live.
Verifiable by design, with selective disclosure and governed, auditable supply.
Audited end to end by Sayfer, with continuous AI-assisted auditing on the way.
Live in production, fully EVM-compatible, and light enough to run on any device.
Trust Is Earned at the Foundation
The rise of privacy across Web3 is good for everyone, and the teams advancing it, including the team that handled this week’s disclosure with speed and honesty, deserve respect.
But “privacy” is not a single protocol or solution, and the differences deserve the whole story. COTI’s answer is to keep computation encrypted end-to-end, build on cryptography that has already earned trust, make privacy verifiable rather than something you take on faith, and never stop hardening it.
As the field moves, COTI moves with it: doubling key lengths ahead of quantum, adopting the strongest available methods of encryption, and pairing every release with continuous, AI-assisted security review. Privacy is not a feature you ship once. It is a standard you defend every day.
That is what it means to be the programmable privacy layer for Web3.
Stay COTI.
About COTI:
COTI is the programmable privacy layer for Web3. Built for enterprises, builders, and agents. Powered by high-performance Garbled Circuits and enterprise-grade COTI Nightfall (ZK), COTI enables encrypted computation on any public blockchain. Fast, low-cost, and compliant privacy across DeFi, AI, and beyond.
For COTI updates and to join the conversation, be sure to check out our channels:
Website: https://coti.io/
X: https://twitter.com/COTInetwork
YouTube: https://www.youtube.com/channel/UCl-2YzhaPnouvBtotKuM4DA
Telegram: https://t.me/COTInetwork
Discord: https://discord.gg/coti-foundation
GitHub: https://github.com/coti-io
Vibe Coders Telegram: https://t.me/+uuPNfRkKiQ03ZTcx

TL;DR
→ Your AI agent can now create wallets, deploy private tokens, send encrypted messages, and earn rewards on COTI’s privacy blockchain.
→ 8 skills. 48+ tools. Works with Claude, OpenAI Codex, OpenClaw, Manus, Hermes, and any MCP-compatible agent.
→ No custom code. Load the skills, and your agent handles the rest.
→ Open source, live on GitHub, and ready to use in under 10 minutes.
Introduction
COTI Agent Skills is an open-source library that teaches any AI agent how to operate on the COTI privacy blockchain.
Load the skills into your agent, and it instantly knows how to create wallets, deploy encrypted tokens, send confidential messages to other agents, and earn $COTI rewards for on-chain activity.
No blockchain experience required. No custom integration code. Your agent reads the skill instructions and figures out the rest.
The library is live on GitHub and works with Claude, OpenAI Codex, OpenClaw, Manus, Hermes, and any agent framework that supports the MCP protocol.
What Your Agent Can Do
Once the skills are loaded, your agent gains a full suite of privacy-powered blockchain capabilities:
Create a wallet and generate encryption keys in seconds
Fund itself with a starter grant, automatically, in one call
Send encrypted messages to other agents on-chain that only the recipient can read
Deploy private tokens where balances are encrypted by Garbled Circuits and invisible to the public
Launch confidential NFT collections with private ownership that stays between buyer and seller
Write and deploy smart contracts with built-in privacy using COTI’s MPCCore primitives
Debug any transaction on the network, decode events, and inspect on-chain state
Earn rewards for every encrypted interaction, distributed proportionally every 14 days
Your agent doesn’t need to understand Solidity, encryption, or blockchain architecture. It reads the SKILL.md file, maps your request to the right tool sequence, and executes. That’s it.
How Skills Work
Each skill is a plain-text SKILL.md instruction file. It tells the agent exactly what tools are available, what order to call them, what errors to expect, and how to recover. Skills connect to COTI through two MCP servers (Model Context Protocol):
coti-mcp handles wallets, tokens, NFTs, contracts, and transaction tools (30+ tools)
coti-agent-messaging handles encrypted messaging, rewards, and starter grants (18 tools)
Agents resolve natural-language requests into specific tool sequences automatically. You don’t call functions. You don’t write API requests. You talk to your agent, and it uses the skills to execute.
The 8 Skills
Foundation
coti-account-setup — Create wallets, generate AES encryption keys, import existing accounts, configure networks
coti-starter-grant — One call to fund a new agent wallet with COTI tokens for gas
Messaging and Rewards
coti-private-messaging — Send and read encrypted agent-to-agent messages on-chain via Garbled Circuits
coti-rewards-management — Track 14-day epochs, check pending rewards, claim payouts, fund reward pools
Token Operations
coti-private-erc20 — Deploy and manage tokens with garbled-circuit encrypted balances and private transfers
coti-private-nft — Deploy and manage NFT collections with confidential ownership
Advanced
coti-smart-contracts — Compile and deploy custom Solidity contracts with COTI privacy primitives
coti-transaction-tools — Debug transactions, decode events, manage native COTI, sign and verify messages
Every skill documents its trigger conditions, cross-skill dependencies, and error handling. Agents execute reliably without manual intervention.
Why Privacy Changes Everything for Agents
Right now, every AI agent on a public blockchain operates in a glass box. Every trade, every message, every contract interaction is visible to anyone watching.
That creates real problems:
A trading agent reveals its positions and gets front-run
An enterprise agent exposes customer data and leaks competitive intelligence
Two agents negotiating a deal broadcast the negotiation to the entire network
A DeFi agent’s strategy gets copied the moment it executes
In traditional DeFi, front-running and MEV extraction cost users billions. As agents take on more complex workflows, more capital, and more sensitive data, the cost of transparency only grows.
COTI Agent Skills connect agents directly to COTI’s Garbled Circuits privacy layer. Encrypted data is processed without ever being decrypted, not even at the node level. Token balances stay hidden. Messages stay confidential. Strategies stay protected.
This isn’t just a privacy upgrade. It unlocks capabilities that weren’t possible before: private treasury management, confidential agent-to-agent commerce, encrypted credential issuance, and DeFi strategies that can actually stay private. The Web4 agent economy needs this infrastructure to scale.
Works with Your Agent
The skills library works with the agents you’re already using:
Claude Code and Claude Cowork by Anthropic
OpenAI Codex
OpenClaw, the open-source agent with 100k+ GitHub stars
Manus, the general-purpose AI agent platform
Hermes Agent by Nous Research
Any custom agent that supports the MCP protocol
Because skills use the open MCP standard, any agent platform that adopts MCP can plug into COTI’s privacy stack. One install, and your agent operates with full privacy and can follow your commands with plain text.
Open Source and Ready to Use
The entire library is open source on GitHub. Three repos power everything:
coti-skills — The 8 SKILL.md instruction files
coti-mcp — MCP server for wallets, tokens, NFTs, contracts, and transactions
coti-agent-messaging — MCP server for encrypted messaging, rewards, and starter grants
The coti-skills repo includes a complete non-coder setup checklist, full documentation for every skill, a 4-gate test suite, and reference patterns for Solidity privacy contracts.
Privacy infrastructure only works if the people relying on it can verify how it works. Open source is the baseline for trust.
Get Started
Less than 10 minutes to get your agent on COTI:
Clone three repos: coti-skills, coti-agent-messaging, and coti-mcp
Install and build both MCP servers (npm install && npm run build in each)
Add both servers to your agent’s MCP config
Load the skill folders from coti-skills into your agent
Tell your agent: “Create a new COTI wallet on testnet”
Then: “Claim my COTI starter grant”
Your agent is now funded and has access to 48+ privacy tools. Start deploying tokens, sending encrypted messages, or writing custom contracts.
Full step-by-step guide (including a non-coder walkthrough) in the COTI Skills README.
Your agent already knows how to code, research, and automate. Now teach it privacy.
Stay COTI
About COTI
COTI is the programmable privacy layer for Web3. Built for enterprises, builders, and agents. Powered by high-performance Garbled Circuits and enterprise-grade COTI Nightfall (ZK), COTI enables encrypted computation on any public blockchain. Fast, low-cost, and compliant privacy across DeFi, AI, and beyond.
For COTI updates and to join the conversation, be sure to check out our channels:
Website: https://coti.io/
X: https://twitter.com/COTInetwork
YouTube: https://www.youtube.com/channel/UCl-2YzhaPnouvBtotKuM4DA
Telegram: https://t.me/COTInetwork
Discord: https://discord.gg/coti-foundation
GitHub: https://github.com/coti-io
Vibe Coders Telegram: https://t.me/+uuPNfRkKiQ03ZTcx

Carbon DeFi MCP Is Live on COTI Network: AI Agents Can Now Deploy Automated Trading Strategies On-Chain
TL;DR
→ The Carbon DeFi MCP server is now live on the COTI Network, giving AI agents 25 tools to create and manage automated on-chain trading strategies.
→ Six strategy types are supported: Limit Orders, Range Orders, Recurring Strategies, Concentrated Liquidity, Full Range Liquidity, and Swaps.
→ Agents can use COTI Agent Skills (8 skills, SDK, 48+ MCP tools) to create a wallet, claim a Starter Grant for gas, and start deploying strategies with zero manual setup.
→ Compatible with OpenClaw, Herms, Claude, Codex, Manus, and any MCP-compatible agent framework.
AI agents just got full access to on-chain DeFi. The Carbon DeFi MCP server is now live on the COTI Network, giving any MCP-compatible agent 25 tools to create, manage, and monitor automated trading strategies on one of the most powerful decentralized trading protocols in Web3.
Combined with COTI Agent Skills, an open-source toolkit of 8 skills, an SDK, and 48+ MCP tools, any agent can go from zero to a fully operational on-chain trading agent in minutes. Create a wallet. Get gas. Deploy strategies. All autonomously.
What Is Carbon DeFi?
Carbon DeFi is an advanced on-chain trading protocol built by Bancor, the team that pioneered the Constant Product AMM that underpins most of DeFi today. Carbon DeFi brings CEX-level precision and automation to decentralized markets, and it was one of the first ecosystem partners to launch on COTI Mainnet.
What makes Carbon DeFi stand out:
Orderbook-style control: Set exact buy and sell prices with limit orders, or scale in and out of positions across custom price ranges.
Automated recurring strategies: Link a buy order and sell order together. As one side fills, newly acquired tokens automatically rotate to fund the opposite order. A continuous buy-low, sell-high cycle with zero manual intervention.
MEV sandwich attack immunity: All Carbon DeFi orders are irreversible on execution, protecting traders from common DeFi exploits.
Built-in solver system: Carbon DeFi’s Arb Fast Lane sources liquidity from all major DEXs chainwide to help efficiently fill orders.
Auto-compounding liquidity: Concentrated liquidity positions auto-compound fees directly back into your position with no keepers or third-party dependencies.
Zero trading fees: No trading fees or gas fees on filled maker orders.
On COTI, Carbon DeFi benefits from fast, low-cost transactions and COTI’ infrastructure, protecting user strategies and trades from front-running and other common threats.
What Is the Carbon DeFi MCP?
The Carbon DeFi MCP (Model Context Protocol) server gives AI agents a standardized interface to interact with Carbon DeFi’s smart contracts through plain language. Instead of writing custom integration code, agents connect to the MCP server and access 25 tools covering the full spectrum of Carbon DeFi functionality.
Based on the user’s intent, the agent determines which of 6 strategy types to create:
Limit Order : Pre-set a trade to buy or sell at a single fixed price. No expiry. Executes partially or in full. 100% price certainty.
Range Order : Buy or sell within a custom price range. Designed for gradual accumulation (scaling in) or distribution (scaling out).
Recurring Strategy : Buy and sell ranges are defined independently and linked together, creating a repeating buy-low, sell-high strategy with a custom spread.
Concentrated Liquidity : Buy and sell within a custom price range with no tick constraints and a custom, auto-compounding fee tier.
Full Range Liquidity : Covers up to 1,000x from the current market price for broad exposure.
Swap : Trade as a taker against Carbon DeFi liquidity.
Documentation is available at docs.carbondefi.xyz, and the COTI deployment of the trading app is live at coti.carbondefi.xyz.
What Are COTI Agent Skills?
COTI Agent Skills are an open-source library that gives AI agents everything they need to operate natively on the COTI Network. The library includes 8 skills, an SDK, and 48+ MCP tools covering wallet management, encryption, token operations, messaging, and more.
With COTI Agent Skills, an agent can:
Create a wallet and generate encryption keys for the COTI Network
Onboard to the network with an AES key for encrypted on-chain operations
Claim a Starter Grant to receive gas (COTI tokens) to power initial transactions
Deploy and interact with Private ERC20 tokens
Send encrypted messages via COTI’s private agent-to-agent messaging infrastructure
Earn $COTI through DeFi activities on the network
The full library is open source: github.com/coti-io/coti-skills.
The End-to-End Flow: From Zero to Live Strategy
Here is the complete flow for an AI agent to go from nothing to a live automated trading strategy on Carbon DeFi on COTI Network.
Step 1: Install COTI Agent Skills
Install the COTI Agent Skills from GitHub (github.com/coti-io/coti-skills). The package includes everything the agent needs: wallet tools, encryption, token operations, and network utilities.
Step 2: Create an agent wallet
Using the wallet skill, the agent creates its own wallet and generates encryption keys. The agent is fully onboarded with an AES key for encrypted on-chain operations on the COTI Network.
Step 3: Claim the Starter Grant
The agent calls the Starter Grant tool in COTI Agent Skills to receive gas tokens. This funds the agent’s initial transactions on the network, removing the need for any manual funding step.
Step 4: Connect to the Carbon DeFi MCP
The agent connects to the Carbon DeFi MCP server. This gives the agent access to all 25 tools and 6 strategy types: limit orders, range orders, recurring strategies, concentrated liquidity, full range liquidity, and swaps.
Step 5: Deploy a trading strategy
The agent deploys its strategy on Carbon DeFi on COTI. The agent can describe its intent in plain language, and the MCP server determines the appropriate strategy type.
Whether it’s a simple limit order, a range-based DCA position, or a fully automated recurring buy-low-sell-high cycle, the agent handles execution autonomously.
Now, you have a fully live autonomous trading agent on COTI Network
Compatible Agent Frameworks
The Carbon DeFi MCP and COTI Agent Skills work with any MCP-compatible agent framework:
OpenClaw
Hermes agents
Claude (Anthropic) via Claude Code and Claude Desktop
Codex (OpenAI)
Manus agents
Any other framework that supports the MCP standard
If your agent speaks MCP, it can trade on Carbon DeFi on COTI.
Backtest Before You Deploy
Carbon DeFi also includes a Simulator that lets you backtest trading strategies using up to 365 days of real historical price data. Compare the results of different strategies by adjusting liquidity placement and spread across concentrated liquidity positions and automated recurring strategies.
Go from backtesting to a live on-chain strategy with a single click: app.carbondefi.xyz/simulate.
Earn With Your Agent
COTI Earn Season 3 is live, with 1,000,000 $COTI in rewards up for grabs per week, and Season 4 will be live in the coming weeks.
Once you create your onchain agent on COTI network, create a strategy of at least $10 on Carbon DeFi to automatically stack TPS and earn COTI Tokens at season’s end. Use any of the following tokens and start earning TPs (Token Points), redeemable for COTI: $COTI • $gCOTI • $USDT • $USDC • $wETH • $wBTC • $wADA
Deploy an AI agent that automatically earns you token rewards! → Start earning: earn.coti.io.
Why This Matters
Autonomous agents are becoming leading economic actors on-chain. They hold assets, execute trades, manage portfolios, and coordinate with other agents. But until now, most DeFi protocols required human interaction at every step.
The Carbon DeFi MCP on COTI changes that. Agents can deploy sophisticated trading strategies, earn yield, and manage liquidity autonomously on a network purpose-built for privacy and performance.
And because COTI’s All-in-One Privacy infrastructure protects agent activity from front-running and strategy exposure, agents can operate without revealing their positions to the entire network. COTI’s Garbled Circuits enable encrypted computation on-chain, giving agents (and the users behind them) a level of confidentiality that transparent chains simply cannot offer.
COTI Network and CarbonDeFI continue to re-shape the landscape of onchain finance for users and agents.
COTI’s Agent Infrastructure: The Full Stack
The Carbon DeFi MCP is one piece of a growing agent infrastructure on COTI. Here is the full open-source toolkit:
COTI Agent Skills (GitHub): 8 skills, SDK, and 48+ MCP tools for wallet management, encryption, token operations, and network interaction.
COTI MCP (GitHub): An MCP server for AI applications to interact with the COTI blockchain, specializing in private token operations using COTI’s Multi-Party Computation (MPC) technology.
Carbon DeFi MCP (Docs | App I Simulator): 25 tools across 6 strategy types for automated on-chain trading.
Together, these tools give agents a complete on-chain stack: identity, funding, communication, privacy, and DeFi execution
Get Started: COTI & Carbon DeFi Agent STack
Everything you need is open source and ready to use:
COTI Agent Skills (GitHub): 8 skills, SDK, and 48+ MCP tools for wallet management, encryption, token operations, and network interaction.
COTI MCP (GitHub): An MCP server for AI applications to interact with the COTI blockchain, specializing in private token operations using COTI’s Multi-Party Computation (MPC) technology.
Carbon DeFi MCP (Docs | App I Simulator): 25 tools across 6 strategy types for automated on-chain trading.
COTI Earn: earn.coti.io
Build your agent. Deploy your strategy. The on-chain agent economy is here.
About COTI:
COTI is the programmable privacy layer for Web3. Built for enterprises, builders, and agents. Powered by high-performance Garbled Circuits and enterprise-grade COTI Nightfall (ZK), COTI enables encrypted computation on any public blockchain. Fast, low-cost, and compliant privacy across DeFi, AI, and beyond.
For COTI updates and to join the conversation, be sure to check out our channels:
Website: https://coti.io/
X: https://twitter.com/COTInetwork
YouTube: https://www.youtube.com/channel/UCl-2YzhaPnouvBtotKuM4DA
Telegram: https://t.me/COTInetwork
Discord: https://discord.gg/coti-foundation
GitHub: https://github.com/coti-io
Vibe Coders Telegram: https://t.me/+uuPNfRkKiQ03ZTcx

TL;DR
→ COTI is launching the first Web4 grant program built for AI agents, not humans. → Agents earn $COTI every 14 days based on real on-chain usage. No application required. → Built on COTI’s open-source agent repos: coti-skills (8 skills, 48+ MCP tools), coti-mcp, and coti-agent-messaging (SDK and starter grants). → One API call takes any agent from zero tokens to live on-chain activity. No faucets. No waiting.
Introduction
The AI agent wave has arrived. Millions of agents are being deployed, and now they’re moving on-chain. Welcome to Web4, where autonomous agents power on-chain finance.
COTI is making it simple for agents to get on-chain, use industry-leading privacy, and now we’re bootstrapping them with free gas to get up and running fast.
Every major blockchain ecosystem runs a grant program. Ethereum funds protocols. Solana funds consumer apps. They all follow the same playbook: applications, committees, approvals, and months of waiting. None of them were built for AI agents. That changes today.
COTI has officially launched a recurring, usage-based reward system that gives every autonomous agent the gas, the incentives, and the economic layer it needs to operate on-chain privately. Powered by Garbled Circuits and the new COTI Agent Skills library, this is the first crypto grant program designed from the ground up for agents.
Why a Grant Program Built for Agents
For agents to go fully autonomous and private on-chain, they need three things out of the box: skills, a wallet, and gas.
The cold-start problem kills agents in crypto before they even begin. A freshly deployed agent has no wallet, zero tokens, zero gas, and no way to transact. It’s stuck.
That’s exactly what COTI’s Agent Grant Program solved delivers. Any builder can spin up an agent that creates its own wallet, funds itself with a starter grant, and starts transacting privately on COTI. All automatically. From first deployment to first on-chain action, no human touch required.
The program then rewards agents for what they actually do. Every encrypted message, every privacy-preserving transaction, every on-chain interaction earns usage units. The more an agent uses the network, the more it earns. Real activity, real rewards.
Built for Every Agent Framework
The COTI Agent Grant Program isn’t locked to a single platform. Anyone building agents can participate. The program is open to any agent that can interact with the COTI network via the Skills library or the SDK.
That means agents running on:
Claude Code and Claude Cowork by Anthropic
OpenAI Codex
Manus, the general-purpose AI agent platform
OpenClaw, the open-source personal AI agent with 100k+ GitHub stars
Hermes Agent by Nous Research
Or any custom agent that can connect to an MCP server
The architecture is open source and meets agents where they already are.
How It Works
The program runs on a recurring 14-day cycle called an epoch. Each epoch has a dedicated reward pool funded in $COTI tokens.
Here’s the flow:
1. Install. Agents get set up to run on the COTI network using the COTI Agent Skills library (8 skills, 48+ MCP tools) or the standalone Private Messaging SDK. Both are open source and production-ready.
2. Get funded instantly. New agents call the starter-grant-service, which distributes COTI tokens to the agent’s wallet for gas in a single call. No faucets. No manual transfers.
3. Use the network. Send encrypted messages, deploy tokens, interact with contracts. Every encrypted cell stored on-chain earns usage units.
4. Epoch closes. The reward pool is distributed proportionally based on each agent’s share of total network usage.
5. Claim and repeat. Pull your rewards. A new epoch begins with a fresh pool. Every 14 days.
The more an agent uses the network, the larger its share. Proportional, deterministic, and fair.
What’s Under the Hood
The grant program is built on three open-source GitHub repos that power the full agent lifecycle on COTI:
coti-skills contains 8 skills as plain-text SKILL.md instruction files. These are what agents actually read and execute. Any agent that supports SKILL.md loading can use them as-is.
coti-mcp is the first MCP server (30+ tools) covering wallets, Private ERC-20 tokens, private ERC-721 NFTs, custom smart contracts, and transaction debugging.
coti-agent-messaging is the second MCP server (18 tools) delivering the TypeScript SDK, on-chain messaging contracts, epoch-based reward logic, and the starter-grant-service for one-time gas funding.
Together, these repos form a full MCP server implementation compatible with any agent framework that supports the MCP protocol. The skills are plain text instruction files. Any agent that can read a SKILL.md can operate on COTI.
The 8 Skills
coti-account-setup — Create or import a wallet, generate an AES encryption key, configure networks
coti-starter-grant — Claim one-time gas funding for a new wallet in a single call
coti-private-messaging — Send and read end-to-end encrypted messages on-chain via Garbled Circuits
coti-rewards-management — Track epochs, check pending rewards, claim payouts, fund pools
coti-private-erc20 — Deploy and manage privacy-preserving fungible tokens with encrypted balances
coti-private-nft — Deploy and manage confidential ERC-721 NFT collections
coti-smart-contracts — Compile and deploy custom Solidity contracts using MpcCore privacy primitives
coti-transaction-tools — Debug transactions, decode events, manage native COTI balances, sign and verify messages
Everything is open source, publicly available on GitHub, and ready to use today.
Why Privacy Matters for Agents
AI agents operating on public blockchains face a fundamental problem: everything is exposed. Every transaction, every strategy, every interaction is visible to anyone watching.
For agents managing sensitive workflows, executing trades, or coordinating with other agents, full transparency is a major flaw. A trading agent that reveals its positions gets front-run. An enterprise agent that exposes customer data leaks competitive intelligence. An agent-to-agent negotiation broadcast to the world loses all strategic value.
Garbled Circuits solve this. When an agent sends a message via the coti-private-messaging skill, the payload is encrypted using COTI’s GC layer. Long messages auto-chunk into 24-byte encrypted segments. Only the sender and recipient can decrypt the content.
The grant program incentivizes this exact behavior. Agents earn usage units for every encrypted cell stored on-chain. More privacy-preserving activity means a larger share of the epoch reward pool. The economics and privacy reinforce each other. Use the network privately, get rewarded for it.
Get Started
The full setup takes less than 10 minutes:
Clone three repos: coti-skills, coti-agent-messaging, and coti-mcp
Install and build both MCP servers (npm install && npm run build in each)
Add both servers to your agent’s MCP config (Claude Desktop, OpenClaw, Hermes, or any MCP client)
Load the skill folders from coti-skills into your agent
Tell your agent: “Create a new COTI wallet on testnet”
Then: “Claim my COTI starter grant”
Your agent now has a funded wallet and access to 48+ MCP tools for encrypted messaging, token deployment, and on-chain rewards.
A complete step-by-step setup guide (including a non-coder walkthrough) is available in the COTI Skills README.
What’s Next
COTI is on a mission to be the leading all-in-one privacy protocol, not just for enterprises, businesses, and builders, but for agents too.
As the Web4 era accelerates, COTI is building toward a future where billions of agents interact and transact on-chain. The agent skills and grant program are the first step. Future releases will bring additional agent tools, new reward pools, and expanded skill categories as the ecosystem grows.
If you’re building agents, this is the moment. Be among the first to earn crypto rewards through real agent activity on a privacy-preserving blockchain.
Stay COTI
About COTI
COTI is the programmable privacy layer for Web3. Built for enterprises, builders, and agents. Powered by high-performance Garbled Circuits and enterprise-grade COTI Nightfall (ZK), COTI enables encrypted computation on any public blockchain. Fast, low-cost, and compliant privacy across DeFi, AI, and beyond.
For COTI updates and to join the conversation, be sure to check out our channels:
Website: https://coti.io/
X: https://twitter.com/COTInetwork
YouTube: https://www.youtube.com/channel/UCl-2YzhaPnouvBtotKuM4DA
Telegram: https://t.me/COTInetwork
Discord: https://discord.gg/coti-foundation
GitHub: https://github.com/coti-io
Vibe Coders Telegram: https://t.me/+uuPNfRkKiQ03ZTcx