Dataline has raised $10.6 million to expand a decentralized marketplace where data publishers sell feeds to AI agents and get paid per API call. Lattice and Samsung-backed Dispersion Capital co-led the round, with Symbolic Capital participating.
The funding lands during a busy stretch for agent-infrastructure deals. Investors are placing bets on the pipes that feed AI agents, not just the agents themselves.
How the marketplace works
Agents pay through the x402 protocol, a payment standard built for machine-to-machine transactions. Each response carries a confidence score calculated from three inputs: source reliability, freshness and consistency.
The company says the system has powered 19.4 million on-chain transactions and that its community spans more than 300,000 members globally. Builders on Base and Robinhood Chain get complimentary access through a Launch Partners cohort, receiving API calls with AI-graded confidence scores for normalized market data.
Co-founder Nuo Shi said the company built agents before it built the marketplace. "After building AI agents ourselves, we realized that clean and reliable data is the lifeblood of AI," Shi said.
A confidence score for every data pull
The scoring model matters for anyone whose work depends on agent output. A confidence score attached to each response gives agents a way to weigh a stale or inconsistent source before acting on it. That's a different approach from buying data in bulk and hoping the feed holds up.
Per-call pricing also changes the cost math. Teams pay for what their agents actually query rather than committing to an annual data contract. For budget owners comparing vendors, that shifts spend from fixed line items to usage-based costs that scale with agent activity.
Professionals building or buying agent workflows can find structured AI Agent Courses covering automation and tool use.
Why this matters for finance, insurance, marketing and sales teams
If your team runs agents on market data, lead lists or customer records, the confidence score is the part to watch. It gives you a filter for agent output before it reaches a quote, a claim file or a campaign. Ask vendors whether their data feeds carry a comparable reliability signal, and test how your agents behave when that score is low. The funding suggests more suppliers will start selling data this way, which means procurement teams will soon be comparing per-call pricing against traditional subscriptions.
The round was originally reported by PYMNTS.