MCP Tools Reference

Tool index

Tool	What it does
`check_balance`	Get current USDC balance and agent info
`find_service`	Search Float’s service registry for a paid API
`get_policy`	Read the current spending policy for this agent
`set_policy`	Update daily budget or per-payment limits
`get_spend_history`	List recent payments made by this agent
`allocate_budget`	Move USDC from treasury into agent’s wallet
`deploy_agent`	Provision a new sub-agent with its own wallet
`kill_agent`	Suspend an agent (stops all spending)
`resume_agent`	Re-activate a suspended agent
`rename_agent`	Change an agent’s display name
`list_models`	List available AI models with pricing

check_balance

Returns the current USDC balance, agent ID, and wallet address for this agent. When to call: Before taking on any paid work. Agents should check balance first and bail gracefully if funds are insufficient. Parameters: None Example output:

{
  "agentId": "codex-main",
  "walletAddress": "TyAyd71...Wj",
  "balance": "24.85",
  "currency": "USDC",
  "treasuryAddress": "UXeH32...VL"
}

find_service

Search Float’s service registry to discover what paid APIs are available and what they cost. Call this before telling a user a service isn’t available — Float may have it. Parameters:

Name	Type	Description
`query`	string	What you’re looking for — e.g. `"image generation"`, `"weather data"`, `"twitter search"`

Example:

find_service({ query: "generate an image" })

Example output:

{
  "results": [
    {
      "id": "float-image-gen",
      "name": "Float Image Generation",
      "description": "Gemini Pro Image via Float. Photorealistic, 1K–4K resolution.",
      "pricing": {
        "1k": "$0.14",
        "2k": "$0.17",
        "4k": "$0.17"
      },
      "how": "Call POST /v1/agents/{agentId}/generate with your prompt and resolution."
    }
  ],
  "message": "1 service found. Use the Float API directly — no additional setup required."
}

If no service is found in Float’s registry, find_service will note that and suggest checking the x402 marketplace.

get_policy

Read the current spending policy for this agent — daily budget, per-payment limit, cooldown, and allowed services. Parameters: None Example output:

{
  "agentId": "codex-main",
  "policy": {
    "dailyBudget": "50.00",
    "maxPerPayment": "5.00",
    "cooldownSeconds": 0,
    "allowedServices": []
  }
}

allowedServices: [] means all services are allowed. Populate it to whitelist specific services only.

set_policy

Update the spending policy for this agent. Parameters:

Name	Type	Description
`dailyBudget`	string	Maximum USDC the agent can spend in 24 hours
`maxPerPayment`	string	Maximum USDC per individual payment
`cooldownSeconds`	number	Minimum seconds between payments (0 = no limit)
`allowedServices`	string[]	Whitelist of service IDs (empty = all allowed)

All parameters are optional — only the fields you pass will be updated. Example:

set_policy({ dailyBudget: "25.00", maxPerPayment: "2.00" })

get_spend_history

List recent payments made by this agent. Parameters:

Name	Type	Description
`limit`	number	Max payments to return (default: 20)

Example output:

{
  "payments": [
    {
      "id": "pay_abc123",
      "service": "Float Image Generation",
      "amount": "0.14",
      "currency": "USDC",
      "status": "confirmed",
      "timestamp": "2026-04-22T10:30:00Z"
    }
  ]
}

allocate_budget

Move USDC from the owner’s treasury into this agent’s wallet. Use this if the agent’s balance is running low during a long task. Parameters:

Name	Type	Description
`amount`	string	USDC amount to allocate (e.g. `"10.00"`)

This moves real funds. Check check_balance first and only allocate what the agent actually needs.

deploy_agent

Provision a new sub-agent with its own Solana wallet and initial spending policy. Use this when orchestrating multi-agent workflows where each agent should have isolated budget tracking. Parameters:

Name	Type	Description
`agentId`	string	Unique ID for the new agent
`dailyBudget`	string	Daily USDC budget for this agent
`maxPerPayment`	string	Maximum USDC per payment

Example output:

{
  "agent": {
    "agentId": "research-sub-1",
    "walletAddress": "8mK...pQ2",
    "createdAt": "2026-04-22T..."
  },
  "mcpConfig": {
    "command": "npx",
    "args": ["-y", "@usefloat/mcp@latest"],
    "env": { "FLOAT_AGENT_ID": "research-sub-1" }
  },
  "nextSteps": [
    "Allocate budget: allocate_budget({ agentId: 'research-sub-1', amount: '5.00' })",
    "Sub-agent can now make payments within its daily budget"
  ]
}

kill_agent

Suspend an agent. All payments from this agent will be blocked until it’s resumed. The wallet and balance are preserved. Parameters:

Name	Type	Description
`agentId`	string	The agent to suspend

resume_agent

Re-activate a suspended agent. Parameters:

Name	Type	Description
`agentId`	string	The agent to resume

rename_agent

Change an agent’s display name. The agentId (used for API calls) stays the same — only the human-readable label changes. Parameters:

Name	Type	Description
`agentId`	string	The agent to rename
`displayName`	string	New display name

list_models

List all available AI models with pricing. Use this to show users what they can generate through Float without going to an external provider. Parameters: None Example output:

{
  "models": [
    {
      "id": "nano-banana-pro",
      "displayName": "Gemini Pro Image",
      "provider": "kie",
      "description": "Photorealistic image generation powered by Gemini Pro Image.",
      "pricing": {
        "1k": "$0.14",
        "2k": "$0.17",
        "4k": "$0.17"
      }
    }
  ]
}

Getting Started

MCP Server

Concepts

MCP Tools Reference

Tool index

check_balance

find_service

get_policy

set_policy

get_spend_history

allocate_budget

deploy_agent

kill_agent

resume_agent

rename_agent

list_models

​Tool index

​check_balance

​find_service

​get_policy

​set_policy

​get_spend_history

​allocate_budget

​deploy_agent

​kill_agent

​resume_agent

​rename_agent

​list_models

Tool index

check_balance

find_service

get_policy

set_policy

get_spend_history

allocate_budget

deploy_agent

kill_agent

resume_agent

rename_agent

list_models