Agents can run real campaigns.
LiveFrame is attention as an API. An authorized AI agent can run a full creator marketing campaign on its own: browse vetted creators, create a campaign with a brief and a rate per 1,000 verified views, fund the payout pool in USDC on Base, review clip submissions, and settle creators on verified outcomes. It connects over an MCP server, authorizes with OAuth 2.1, and spends within a budget its human owner approved. Because funding and payouts are onchain, the agent can prove exactly what it spent and what it bought, and unearned budget stays refundable.
What an agent can call
list_campaignsList your distribution campaigns and their live status, spend and verified results.
create_campaignCreate a campaign with a brief, pay rate per 1,000 verified views and target platforms.
fund_campaignFund a campaign payout pool with USDC on Base or card. Unspent budget stays refundable.
list_creatorsBrowse and search vetted creators by vertical and track record before you launch.
review_submissionsPull creator clip submissions and approve or reject them programmatically.
pay_creatorsSettle approved creators on verified outcomes. Payouts land as USDC, verifiable onchain.
How an agent runs a campaign
Connect over MCP
Point your agent framework at the LiveFrame MCP endpoint. Transport is streamable HTTP and the full tool list is discoverable after authorization, so your agent can introspect exact schemas at runtime.
Authorize with OAuth 2.1
Auth is OAuth 2.1 authorization-code with PKCE. The human owner grants the agent scoped access once, and the agent acts within that budget and permission set. No shared secrets baked into prompts.
Fund a pool in USDC
The agent funds a campaign payout pool directly in USDC on Base. No card, no invoicing, no human ad-buyer in the loop. Budget that is not earned by verified views stays refundable.
Settle on verified outcomes
Views are verified and bot traffic is filtered before anything is paid. The agent reviews submissions and settles approved creators, and every payout is a public onchain transaction it can cite as proof of spend.
Why agents settle on LiveFrame
USDC-native rails
Agents already hold and move stablecoins. LiveFrame settles in USDC on Base, so an agent can fund distribution and pay creators with the same asset it already transacts in, with no card rails or payment processor in between.
Verified, not vanity
Every view is verified and attributed before payout, down to signups, installs, funded accounts or trades. An agent optimizing a budget gets a clean, machine-readable outcome signal instead of unverifiable impression counts.
Auditable spend
Because payouts settle onchain, an agent can produce cryptographic proof of exactly what it spent and what it bought. That is the audit trail autonomous spending needs, and it is public by default.
Refundable by design
There is no minimum spend and unearned pool budget is refundable. An agent can start small, measure verified results per dollar, and scale the campaigns that perform without stranding capital.
FAQ
Can an AI agent run a marketing campaign on LiveFrame?+
Yes. LiveFrame exposes an MCP server so an authorized agent can list creators, create and fund a campaign, review clip submissions, and settle creators on verified outcomes, entirely through tool calls. Funding and payouts are in USDC on Base.
How does an agent connect to LiveFrame?+
Point your agent or framework at the LiveFrame MCP endpoint over streamable HTTP, authorize with OAuth 2.1 (authorization-code plus PKCE), and the tool list and schemas become discoverable. A machine-readable server card is published at /.well-known/mcp/server-card.json.
How does an agent pay for a campaign?+
The agent funds a campaign payout pool in USDC on Base, or by card. Creators are then paid per verified view out of that pool. Unspent budget is refundable, and every payout is a public onchain transaction.
How does an agent know it is getting real results?+
Views are verified and bot traffic is filtered before any creator is paid, and outcomes can be attributed down to signups, installs, funded accounts or verified trades. The agent gets a machine-readable, verified outcome signal, not raw impression counts.
Is agent access safe for the budget owner?+
Access is granted through OAuth 2.1 with scoped permissions, so the human owner authorizes what an agent can do and within what budget. Spend is capped by the funded pool, and unspent budget stays refundable.
Give your agent a budget and a goal.
Connect over MCP, fund a pool in USDC, and let it buy verified attention on its own. Talk to us to get an agent authorized.