Secure smart wallets, simulation-first execution, gas sponsorship, and typed DeFi actions for AI agents operating across Ethereum, Base, and Arbitrum testnets. One integration from intent to confirmed transaction.
Today's agents reason, plan, and generate. But when it comes to doing - executing a payment, signing a transaction, or calling a contract - they hit a wall.
A full-stack backend infrastructure layer that gives AI agents a secure, programmable identity onchain and the rails to act with it. Already built. Already running.
// One API call. const response = await fetch("https://api.ardentresearch.xyz/execute", { method: "POST", headers: { "content-type": "application/json", "x-api-key": "ak_live_xxx" }, body: JSON.stringify({ agent_id: "research-agent-01", chain: "ethereum", target_contract: "0x1c7D4B...", calldata: "0xa9059cbb...", value: "0" }) }); const result = await response.json(); // result.request_id → 8f4d...c321 // result.status → "queued" // signed webhook confirms final onchain status
Agents work with explicit, protocol-aware actions instead of composing opaque calldata. Every supported write path includes preparation, simulation, and structured execution through the same wallet.
Manage lending positions with health-factor-aware borrowing, debt-aware maximum repayment, and wallet-level asset visibility.
Discover risk-filtered isolated markets, then manage lending, collateral, debt, accrued interest, health, and capacity.
Verify and select the Base USDC or WETH Comet, then inspect positions, current rates, and borrowing capacity.
Prepare and execute exchange, increase, decrease, cancel, and claim actions with market metadata, order reads, and token balances.
Discover the best quoted pool, execute swaps, inspect pool state, and manage liquidity through Balancer's V3 Router and Vault.
Discover the best direct pool, quote, simulate, and execute through the Universal Router with native ETH and explicit hook support.
A single agent identity can operate through chain-specific smart wallets while your integration keeps one API, one authentication model, and one execution workflow.
From agent decision to confirmed onchain transaction with full security, gas management, and auditability built in at every step.
Use the REST API directly, install the zero-dependency CLI, or expose Ardent as tools through MCP. Each path reaches the same typed actions and simulation pipeline.
Get wallet, balance, simulation, execution, and protocol commands in your terminal. The installer also provisions the local files used by supported agent integrations.
curl -fsSL https://raw.githubusercontent.com/Ardent-Ai-Research/agent-execution-platform/master/docs/agent-integration/install.sh | bash
The MCP server turns wallet, simulation, execution, and protocol capabilities into callable tools for compatible desktop AI assistants. The installer detects supported apps and registers the server automatically; you then add your Ardent API key locally.
curl -fsSL https://raw.githubusercontent.com/Ardent-Ai-Research/agent-execution-platform/master/docs/agent-integration/install.sh | bash
Integrate from any language or agent framework using conventional HTTP. Prepare and simulate actions, submit execution requests, and track finality through status endpoints and signed webhooks.
GET https://api.ardentresearch.xyz/wallet?agent_id=my-agent&chain=ethereum
Onchain autonomy should not mean blind execution. Ardent builds inspection, simulation, policy-aware sponsorship, and finality tracking into the transaction lifecycle.
Bundler gas estimation and stateful batch simulation validate the complete account-abstraction path before submission.
Agent signing material is encrypted at rest with AES-256-GCM and is never returned through public API responses.
Transactions can use native funds or qualified paymaster sponsorship with gas recalculated for the final operation.
Protocol adapters validate markets, assets, amounts, and relevant risk constraints before constructing contract calls.
HMAC-SHA256 signed webhooks deliver verifiable execution outcomes with retry handling for transient delivery failures.
Request IDs connect submitted intent, execution status, UserOperations, and confirmed transaction hashes.
A working testnet platform with a real execution engine, encrypted key management, reliable queuing, and focused test coverage.
Agent signing keys persist only as AES-256-GCM encrypted records at rest. HMAC-signed webhooks and rate-limited APIs protect the surrounding execution flow.
One interface across supported chains and protocol adapters. Plug into assistant tools, custom pipelines, or raw API calls.
We are researchers, engineers, and builders at the intersection of AI and blockchain infra. We've lived this problem. We built the solution we needed ourselves.
— Ardent Research
Ardent is live on public testnets and is intended for developers evaluating agent-driven onchain execution before production use.
No. Each agent operates through its own ERC-4337 smart account. Ardent constructs, simulates, signs, and submits authorized operations, but assets remain in the agent's onchain wallet.
Yes. Ardent exposes dedicated simulation endpoints and performs full UserOperation gas estimation in the account-abstraction flow. Protocol actions use the same execution pipeline.
Ethereum Sepolia, Base Sepolia, and Arbitrum Sepolia. Protocol availability follows the deployments listed in the network section above.
Yes. The integration pack includes an MCP server for compatible desktop assistants and an OpenAPI specification for hosted tool integrations such as custom assistant actions.
Not yet. The current public release is a testnet environment for integration testing and product feedback. Do not treat it as a production custody or mainnet execution service.
We're onboarding a limited number of early builders. Apply below and we'll personally send your API key. No queue. No automation. Every applicant reviewed by the founding team.
No automated signup queue. We review every application and deliver your key directly.
Early testers get direct Telegram/Discord line to the founding team throughout testnet.
Your feedback directly influences what we build next. Early builders get priority features.