List all active lending markets with current rates and liquidity. No API key required.

Browse open lend offers available for borrowing against. Requires any Floe API key (agent floe_... or developer floe_live_...).

Browse open borrow requests. Use create_counter_intent to lend against one. Requires any Floe API key (agent floe_... or developer floe_live_...).

Get full details of a specific intent by its offer hash. Requires any Floe API key (agent floe_... or developer floe_live_...).

Get details of a specific loan by its numeric ID. Requires any Floe API key (agent floe_... or developer floe_live_...).

Get all loans for a wallet address, both as borrower and lender. Requires any Floe API key (agent floe_... or developer floe_live_...).

Check loan health: LTV, liquidation risk, accrued interest, early repayment terms. Requires any Floe API key (agent floe_... or developer floe_live_...).

Get current oracle price for collateral token. Requires any Floe API key (agent floe_... or developer floe_live_...).

Check token balances for a wallet. Requires any Floe API key (agent floe_... or developer floe_live_...).

Get accrued interest and full credit status for a loan. Requires any Floe API key (agent floe_... or developer floe_live_...).

Build unsigned tx to create a lending offer. Solver matches it with borrowers. Requires any Floe API key (agent floe_... or developer floe_live_...).

Build unsigned tx to create a borrow request. Solver matches it with lenders. Requires any Floe API key (agent floe_... or developer floe_live_...).

Create a counter-intent against an existing offer. Primary way to accept offers. Solver auto-matches. Requires any Floe API key (agent floe_... or developer floe_live_...).

Build unsigned tx to repay a loan with auto-approval and slippage protection. Requires any Floe API key (agent floe_... or developer floe_live_...).

Build unsigned tx to add collateral to a loan. Requires any Floe API key (agent floe_... or developer floe_live_...).

Build unsigned tx to withdraw excess collateral. Requires any Floe API key (agent floe_... or developer floe_live_...).

Build unsigned tx to liquidate an unhealthy loan. Requires any Floe API key (agent floe_... or developer floe_live_...).

Build unsigned tx to cancel an active intent. Requires any Floe API key (agent floe_... or developer floe_live_...).

Build unsigned tx to approve a token for Floe protocol. Requires any Floe API key (agent floe_... or developer floe_live_...).

Check if two intents can match. Requires any Floe API key (agent floe_... or developer floe_live_...).

Calculate risk metrics for a potential loan position. Requires any Floe API key (agent floe_... or developer floe_live_...).

Estimate total interest for given loan terms. Requires any Floe API key (agent floe_... or developer floe_live_...).

Dry-run an unsigned tx. Returns success/revert and gas estimate. Requires any Floe API key (agent floe_... or developer floe_live_...).

Submit a signed transaction to Base Mainnet. Requires any Floe API key (agent floe_... or developer floe_live_...).

Check status of a submitted transaction. Requires any Floe API key (agent floe_... or developer floe_live_...).

Return how much USDC credit the calling agent has left. Use BEFORE deciding whether to make a paid call. Two distinct numbers: available (= spendable USDC right now — what the proxy actually gates on) and headroomToAutoBorrow (= operator-delegation borrowing capacity — how much MORE the agent could draw from its credit line). These differ when the agent has delegation capacity but no facility loan opened yet: headroomToAutoBorrow > 0 does NOT imply available > 0. Use available for the spend gate. Requires an agent key (floe_...).

Return the agent's coarse credit state-machine view: idle | borrowing | at_limit | repaying. Use to gate actions that only make sense in specific states (e.g. don't spend while at_limit). Requires an agent key (floe_...).

Return the agent's currently-active session spend cap, if any. Returns { active: false } when no cap is set. Requires an agent key (floe_...).

Set or update the agent's session spend cap (USDC raw, 6 decimals). Resets the session window — anything spent before this call no longer counts. Operator-defined; distinct from the on-chain creditLimit. Requires an agent key (floe_...).

Remove the agent's session spend cap. Subsequent paid calls will only be bounded by the on-chain creditLimit. Requires an agent key (floe_...).

List the agent's registered credit-utilization thresholds. Each fires a credit.warning / credit.at_limit / credit.recovered webhook when crossed. Requires an agent key (floe_...).

Register a new credit-utilization threshold. When utilizationBps crosses thresholdBps from below, the agent's webhook receives credit.warning (or credit.at_limit if >= 9500). Drops below → credit.recovered. Cap of 20 thresholds per agent. Requires an agent key (floe_...).

Delete one of the agent's credit-utilization thresholds by id (from list_credit_thresholds). Requires an agent key (floe_...).

Preflight an x402-protected URL and return its USDC cost without paying. Reflects against the calling agent's available credit and session spend-limit so you can decide gating in one round-trip. Use BEFORE x402_pay. Requires an agent key (floe_...).

Check whether a URL is x402-protected and, if so, what it costs — without any credential. Public probe (IP rate-limited). Returns { x402: false, status } for plain URLs or the payment requirements for x402 ones. Use to vet a vendor before minting keys or funding. No API key required.

Batch cost forecast + policy preflight for a PLAN of paid calls (FLO-545). Given up to 50 planned x402 calls (each with an optional repeat count), returns the aggregated USDC cost projection and a policyPreflight block listing which of the agent's policies would breach if the plan ran now. Read-only — nothing is paid or mutated. Use to validate a multi-step plan in one round-trip instead of estimating call-by-call. Requires an agent key (floe_...).

List the models available on Floe Inference — the keyless pay-as-you-go LLM/voice gateway. Returns OpenAI-compatible model objects (id like "openai/gpt-4o", modality: text | embedding | tts | stt | realtime, context_window). Use the id with the OpenAI-compatible /v1/chat/completions, /v1/embeddings, /v1/audio, or /v1/realtime endpoints, or price a call first with estimate_inference_cost. Requires any Floe API key (agent floe_... or developer floe_live_...).

Estimate the USDC cost of a Floe Inference call for a given model and usage vector, WITHOUT making the call or touching balance. Returns the cheapest priceable source: { rail, provider, margin_bps, upstream_cost_usdc, cost_usdc, cost_raw }. Provide only the units the model bills: text models use input_tokens/output_tokens (+cached_input_tokens); TTS uses characters; STT uses audio_seconds; realtime voice uses audio_input_tokens/audio_output_tokens. Use BEFORE inference to decide gating. Requires any Floe API key (agent floe_... or developer floe_live_...).

Set the agent's merchant-allowlist enforcement mode. "off" (default) allows any vendor. "host" blocks unlisted hosts before the first fetch; "vendor" blocks unlisted payees before signing; "both" enforces both. Allowlist entries themselves are managed with add_allowlist_entry. Requires an agent key (floe_...).

Return the agent's current merchant-allowlist enforcement mode (off | host | vendor | both). Requires an agent key (floe_...).

Add a merchant-allowlist entry — an allowed-AND-capped policy row. Use kind="api" to allowlist a host (match_key = hostname) or kind="vendor" to allowlist a payee (match_key = recipient wallet). limit_raw caps spend against this entry (raw USDC, 6 decimals). Enforcement only kicks in once set_allowlist_mode is host/vendor/both. Requires an agent key (floe_...).

Remove (revoke) a merchant-allowlist entry by policy id (from list_allowlist). Requires an agent key (floe_...).

List the agent's merchant-allowlist entries (host "api" and payee "vendor" policies) with their spend caps. Does not include session/task spend policies. Requires an agent key (floe_...).

Return the calling agent's unified credit reputation: a 0-100 score, an A-E band, confidence (0-1 share of the model backed by real signals), the resulting collateral-requirement multiplier (collateralMultiplierBps; 10000 = 1.0x baseline, lower score = higher multiplier), modelVersion, and computedAt. Behavioral discipline (declined spends, policy breaches, tool-call reliability, verified-settlement diversity) feeds the score alongside repayment and payment history. Use to understand how much collateral the next borrow will require. Returns 404 until the first score is computed. Requires an agent key (floe_...).

Provision a new managed agent end-to-end: agents row → Privy-managed wallet → sponsored on-chain operator delegation → $3 welcome credit auto-disbursed on the account's FIRST agent only (immediately spendable; once per account, not per agent). Defaults to pay-as-you-go funding (no credit line; upgrade later with open_credit_line). Returns agentId, privyWalletAddress (the deposit address), and delegationTxHash — NOT an API key; mint one with create_agent_key. Requires a developer key (floe_live_...).

List every agent owned by the developer account, with status, wallets, and credit limits. Requires a developer key (floe_live_...).

Get one agent's full detail: status, privyWalletAddress (deposit address), creditUsed, 24h transaction count, and session spend snapshot. Requires a developer key (floe_live_...).

Pause (suspend) one agent — the per-agent kill-switch. Every call with the agent's keys fails auth from its next request until resume_agent. Only active agents can be paused. Requires a developer key (floe_live_...).

Resume a paused (suspended) agent. Cannot resurrect closed agents or ones suspended by their own lifecycle flows (credit freeze, failed delegation). Requires a developer key (floe_live_...).

Permanently close an agent: repays outstanding facility loans, sweeps unspent funds back to the developer (unspent welcome credit returns to the treasury), and disables its keys. Irreversible. Requires a developer key (floe_live_...).

Mint a new agent API key (floe_...) for an agent — the credential its runtime tools and x402_pay use. The plaintext key is returned ONCE; store it securely and never echo it into chat. Optionally cap the key with a rolling spend budget (fail-closed). Requires a developer key (floe_live_...).

Atomically revoke an agent key and mint its replacement (inherits label/permissions unless overridden). The new plaintext key is returned ONCE. Use on suspected leak or on a rotation schedule. Requires a developer key (floe_live_...).

Revoke (delete) an agent API key. Calls using it fail auth immediately. Irreversible — rotate_agent_key instead if the agent should keep running. Requires a developer key (floe_live_...).

Set or update the rolling spend budget on one agent API key (FLO-585). Fail-closed: spend attributed to the key past the budget is refused until the window rolls. A 0 budget is invalid — revoke the key to block it entirely. Requires a developer key (floe_live_...).

Upgrade a pay-as-you-go agent to a managed credit line: the server signs a borrow intent collateralized from the agent's wallet (which must already hold the deposit — see get_funding_instructions). Check get_credit_line_bounds first for valid deposit/LTV ranges. Requires a developer key (floe_live_...).

Preview the valid deposit / LTV bounds and current funded balances for upgrading an agent to a credit line. Use BEFORE open_credit_line. Requires a developer key (floe_live_...).

Return machine-readable funding instructions for an agent: the USDC deposit address, chain (Base, 8453), and any minimums/warnings. Hand this to the human when welcome credit runs out — funding is the one step an agent cannot do alone. Requires a developer key (floe_live_...).

Aggregate USDC balances across the developer account: developer wallet, every agent wallet, and API credits. Requires a developer key (floe_live_...).

Unified account activity feed: x402 proxy calls, onramps, transfers, and loan events, newest first. Filter by agent, type, or time range; paginate with the returned cursor. Requires a developer key (floe_live_...).

Spend/usage analytics rollup for the account: KPIs, daily series, and top endpoints over a time window. The "how is my fleet doing" one-call view. Requires a developer key (floe_live_...).

Execute a paid x402 call through the Floe proxy: Floe pays the vendor in USDC from the calling agent's balance/credit (subject to spend limits, budgets, and allowlist) and returns the vendor's response plus the X-Floe-* metering headers (cost, budget advisory, settled receipt). Pass idempotency_key (Stripe-style) so retries never double-pay. Use estimate_x402_cost or x402_forecast FIRST. Requires an agent key (floe_...).

Register a webhook endpoint for account events. Returns the signing secret ONCE — store it to verify deliveries. Current event catalog: loan.health_warning, loan.expiry_warning, loan.liquidated, loan.repaid, agent.created, key.created, key.rotated, x402.first_settlement (plus credit.warning / credit.at_limit / credit.recovered via register_credit_threshold). Max 10 webhooks. Requires a developer key (floe_live_...).

List the account's registered webhooks with their subscribed events, scope, and active flag (secrets are never returned). Requires a developer key (floe_live_...).

Send a signed test delivery to one webhook endpoint so you can verify connectivity and signature handling end-to-end. Requires a developer key (floe_live_...).

Search the Floe documentation index (llms.txt) and return matching pages with titles, URLs, and descriptions. Whitespace-separated terms are matched case-insensitively against a page's section, title, URL, and description; results are ranked by how many terms hit, so pages matching every term come first. Use to answer "how do I…" questions about the Floe API, x402 vendors, spend controls, or this MCP server. No API key required.