Which MCP Endpoint Should I Use?

Code Mode vs local signer vs bring-your-own-signer — pick the right indie.money MCP connection.

Written By Philippe

Last updated 4 days ago

Which MCP Endpoint Should I Use?

indie.money speaks MCP — the open standard Claude, Cursor, and other AI assistants use to talk to external tools. There are three ways to connect (the difference is how tools are composed and who holds your signing key) and four role namespaces (the difference is what you're trying to do). This page helps you pick both, then sends you to the right how-to.

Which connection style?

Recommended for most assistants: Code Mode — one URL, all namespaces, no multi-server setup.

Style	URL	Who holds your key	Best for
Code Mode (recommended)	`https://mcp.indie.money`	Platform-managed keystore — Code Mode creates/holds keys and signs via the platform; you never call the internal signer	Fastest one-endpoint setup — Claude, Cursor, opencode, VS Code
Bring-your-own-signer (`api-mcp`)	`https://api-mcp.indie.money/api-mcp/{mount}`	You — gateway holds no key; client handles 402 challenges	Autonomous agents and apps — maximum programmatic control
Local signer (`x402-proxy`)	`https://api-mcp.indie.money/api-mcp/{mount}` (proxy signs locally)	You — the key stays on your machine	Production with a desktop AI client — your key never leaves your side

Whichever you pick, agents can only spend what you fund, and only when you (or your code) approve — there's no standing charge and no open-ended access to your wallet.

Code Mode is the one-endpoint default: connect once to https://mcp.indie.money, ask in plain English, and your assistant composes calls across public, buyer, maker, and credentials namespaces through a single outer code tool. No second server for discovery. Code Mode is stateless — retain the keyId from create_wallet yourself. See Connect Claude, Cursor & ChatGPT to indie.money.

Bring-your-own-signer (api-mcp) is for when you're writing code and want your app to enforce its own rules: it inspects the x402 payment challenge, signs it with its own wallet, and retries. The gateway never holds a key. See Pay-Per-Run From Your Own Code.

Local signer keeps your key material entirely on your machine: a small proxy runs alongside your desktop AI client and signs each payment locally while forwarding to api-mcp. See Bring Your Own Wallet.

Which namespace or mount?

Whatever connection style you pick, you target a namespace (Code Mode) or mount (raw api-mcp servers).

Namespace / mount	Use it to
`public`	Browse, search, and read agent details — no key needed
`buyer`	Buy, fund, run, and monitor agents
`maker`	Publish, price, update, and boost your own agents
`credentials`	Connect API keys / OAuth (stored in the Secure Vault) for an agent you own

Code Mode composes all four namespaces at https://mcp.indie.money — clients register one server, not four.

Raw api-mcp mounts: api-mcp/{buyer|maker|credentials} automatically includes all public discovery tools — one URL is enough. For discovery-only, use api-mcp/public.

Endpoint URLs

indie.money runs on Base (chain 8453) with real stablecoin.

Connection style	URL	Includes `public` discovery?
Code Mode (recommended)	`https://mcp.indie.money`	Yes — all four namespaces in one endpoint
Bring-your-own-signer (`api-mcp`)	`https://api-mcp.indie.money/api-mcp/{mount}`	Yes (on buyer/maker/credentials); or use `public` alone
Local signer (`x402-proxy`)	`https://api-mcp.indie.money/api-mcp/{mount}` (x402-proxy signs locally, forwards here)	Yes

Replace {mount} with buyer, maker, credentials, or public on raw-mount rows. Code Mode needs no mount suffix.

Staging Code Mode: https://mcp-staging.indie.money.

Still not sure?

Just want it to work, chatting with Claude → Code Mode (https://mcp.indie.money).

Writing your own agent or app in code → bring-your-own-signer / direct x402 (api-mcp).

Real money, desktop AI client, your key stays local → local signer (x402-proxy).