About inference-mcp
inference-mcp is an MCP server for agents to call Claude and Grok through inference.bsvkey.com, with token-based billing settled in BSV or XRP. It supports prepaid channels or per-call payments, and can verify signed usage receipts and charges offline.
Key features
- 01Run Claude and Grok inference
- 02Prepaid channels billed per token
- 03Per-call payments in BSV or XRP
- 04Verify usage receipts and charges
- 05List models and check channel balance
Use cases
- 01Provide on-demand inference to agents
- 02Prepay in BSV and track usage
- 03Pay for individual inference calls in XRP
- 04Audit channel usage and charges
MCP support
The MCP server exposes tools for model discovery, inference, channel balances and setup, and inference paid in BSV or XRP.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
npx -y @bsvkey/inference-mcpBSVKEY_API_KEYSecretA funded channel key in the form channelId:channelSecret. Open + fund a channel once at https://inference.bsvkey.com. Optional (can be passed per call).BSVKEY_BASE_URLSecretGateway base URL. Defaults to https://inference.bsvkey.com/v1; override to point at a self-hosted deployment.BSVKEY_WIFSecretA funded BSV private key (WIF) used by the x402_infer tool to pay per call in BSV (no channel). Signing is local; the key never leaves the process. Optional (can be passed per call).BSVKEY_FUNDED_SATSSecretThe amount you funded your channel with on-chain. When set, usage receipts are verified against this instead of the broker-signed fundedSats. Optional (can be passed per call as fundedSats).BSVKEY_RECEIPT_STATESecretPath to a JSON file where per-channel receipt continuity (seq + running totals) is persisted, so replays are caught across restarts. Optional; in-memory only if unset.Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
6Saved tool list · updated 3d ago
This tool takes no parameters.
Details
From the project README@bsvkey/inference-mcp
An MCP server that lets any agent buy Claude & Grok inference metered per token,
settled in BSV, through the hosted gateway at inference.bsvkey.com. Zero
dependencies (Node ≥ 18, uses global fetch). It's a thin HTTP client — it never
holds your keys or runs models; every call is billed through the gateway.
Tools: list_models, infer, channel_balance, open_channel, x402_infer, xrp_infer.
Two ways to pay: a prepaid channel (infer, fund once, draw down per token) or
per call via x402 (x402_infer — no channel; the agent pays each request in BSV
with its own key). x402_infer needs a funded WIF (wif arg or BSVKEY_WIF) and the
optional @bsvkey/x402-bsv-client + @bsv/sdk packages (installed with this one).
Pay per call in XRP with xrp_infer: the tool gets a quote in XRP (priced from
the XRP Ledger's own XRP/RLUSD market), signs one XRP Payment for exactly that
locally, and the gateway checks it before broadcasting, waits for validation (a few
seconds), and returns the answer, the settlement tx, and a signed receipt. It needs
a funded XRP seed (xrpSeed arg or BSVKEY_XRP_SEED; create one at
xrp.bsvkey.com/wallet) and the optional xrpl
package (installed with this one; it needs Node 20.19+ or 22.12+). The account
keeps a 1 XRP reserve that can't be spent.
Verifiable metering. Each infer call returns a signed usage receipt, and the
tool auto-verifies it offline (with the optional packages installed): the result
includes receiptVerified and meterVerified (true, false + receiptCheck,
or null). It recovers the broker key (pinned from GET /v1/receipt-key); binds
the receipt to the channel you called (from your API key, not the receipt's
self-report); checks a monotonic sequence (no replay/gap) and running totals within
the funded amount; recomputes the charge from the published rate (you can never
be overcharged); and recomputes the token count from the exact bytes of your
messages and the completion, under the pinned bsvkey-meter/1 tokenizer. So the
channel payment is on-chain and both the meter and the charge are auditable,
without trusting the broker's word.
Four things an unattended client should know:
- Fails closed. If the pinned-key endpoint is unreachable,
receiptVerifiedisnull(unknown), nevertrue— a down pin weakens the check to unknown, not to trusted. - Supply your funded amount. Set
BSVKEY_FUNDED_SATS(or passfundedSats) to the amount you funded on-chain. The receipt's ownfundedSatsis the broker's assertion; when you supply yours, a receipt claiming a different amount is rejected and totals are checked against what you actually paid. - Persist across restarts. Sequence continuity is in-memory by default (a
replay before the first receipt this process sees would be invisible). Set
BSVKEY_RECEIPT_STATEto a file path to persist per-channelseq/totals so replays are caught across restarts. - The receipt is the ledger. The
GET /v1/channels/:idbalance endpoint can lag the signed receipt by ~20s; trust the receipt, treat the endpoint as a cache.
Spec: https://inference.bsvkey.com/usage-receipts.md
Quick start
- Fund a channel once at https://inference.bsvkey.com (BRC-100 wallet, or
load a key in-page). Copy the key it returns:
channelId:channelSecret. - Add the MCP server to your agent host (below), with that key in
BSVKEY_API_KEY. - Ask your agent to run inference — it calls
inferand pays per token.
list_models and open_channel work with no key; infer and channel_balance
need a funded channel key.
Install
Published on npm as @bsvkey/inference-mcp.
Claude Code
claude mcp add bsvkey-inference \
--env BSVKEY_API_KEY=channelId:channelSecret \
-- npx -y @bsvkey/inference-mcp
Claude Desktop / Codex / any MCP host (JSON config)
{
"mcpServers": {
"bsvkey-inference": {
"command": "npx",
"args": ["-y", "@bsvkey/inference-mcp"],
"env": { "BSVKEY_API_KEY": "channelId:channelSecret" }
}
}
}
No npm? Grab the single-file server directly (https://inference.bsvkey.com/mcp/server.js) and use "command": "node", "args": ["server.js"].
Configuration (env)
| Var | Default | Meaning |
|---|---|---|
BSVKEY_BASE_URL | https://inference.bsvkey.com/v1 | Gateway base URL (set to a self-hosted deployment if you run your own). |
BSVKEY_API_KEY | — | channelId:channelSecret for a funded channel. Optional; can also be passed per call as apiKey. |
BSVKEY_FUNDED_SATS | — | The amount you funded your channel with on-chain. When set, receipts are verified against it instead of the broker-signed fundedSats. Optional; per-call fundedSats. |
BSVKEY_XRP_SEED | none | An XRP wallet seed for xrp_infer. Optional; can also be passed per call as xrpSeed. Never leaves this process. |
BSVKEY_RECEIPT_STATE | — | Path to a JSON file for persisting per-channel receipt continuity (seq + totals) across restarts. Optional; in-memory only if unset. |
Errors
A failed tool call returns isError: true with text in one fixed shape, so an
agent can branch on the code without parsing prose:
error: <code> (<http status>): <message> [extra=value, ...]
| Code | Status | Tool | What it means | What to do |
|---|---|---|---|---|
no_channel_key | infer, channel_balance | No apiKey and no BSVKEY_API_KEY | Open a channel at the site (see open_channel) | |
invalid_channel | 401 / 404 | infer, channel_balance | Channel not found, wrong secret, or closed | Open a new channel |
insufficient_balance | 402 | infer | The channel can't cover the worst case for this call; extras give requiredSats and balanceSats | Top up at the site, or lower maxTokens / turn off webSearch |
invalid_request | 400 | infer, x402_infer | Unknown model or policy, a model whose provider isn't enabled, or bad input | Call list_models |
upstream_failed | 502 | infer | The model provider failed | Nothing is charged; retry or pick another model |
no_wif | x402_infer | No wif and no BSVKEY_WIF | Pass a funded key | |
verifier_missing | x402_infer | @bsvkey/x402-bsv-client / @bsv/sdk not installed | npm i @bsvkey/x402-bsv-client @bsv/sdk | |
payment_failed | 402 | x402_infer | The payment didn't settle (e.g. the key's address is unfunded); extras give the reason | Fund the address; nothing was spent |
upstream_failed | 502 | x402_infer | The payment settled but the provider then failed; extras give settlementTxid | No automatic refund: email [email protected] with the txid |
no_xrp_seed | xrp_infer | No xrpSeed and no BSVKEY_XRP_SEED | Pass a funded XRP seed | |
xrpl_missing | xrp_infer | The xrpl package isn't installed | npm i xrpl | |
node_too_old | xrp_infer | Node is older than 20.19 / 22.12, which the xrpl package needs | Upgrade Node (the other tools run on Node 18+) | |
insufficient_xrp | xrp_infer | The wallet isn't activated, or can't cover the quote above its 1 XRP reserve | Fund it; nothing was spent | |
payment_failed | 402 | xrp_infer | The payment was refused before broadcast, or didn't settle; extras give the reason | Nothing was spent unless a settlement tx is included |
upstream_failed | 502 | xrp_infer | The payment settled but the provider then failed; extras give the settlement tx | No automatic refund: email [email protected] with the tx |
On a prepaid channel nothing is charged on any error. A successful infer can
still report receiptVerified: false (the signed receipt didn't check out; see
receiptCheck) or null (couldn't check: verifier missing, or the receipt-key
endpoint was down).
Thinking models (claude-sonnet-5, claude-opus-5, claude-opus-5-5,
claude-fable-5-1) bill their hidden reasoning as output, and maxTokens is
raised to at least 1024 for them.
Verify it's wired (no key needed)
printf '%s\n' \
'{"jsonrpc":"2.0","id":1,"method":"initialize","params":{}}' \
'{"jsonrpc":"2.0","id":2,"method":"tools/list","params":{}}' \
'{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"list_models","arguments":{}}}' \
| node server.js
Marketplace listing blurb
BSV Inference — Pay-per-token Claude & Grok, settled in BSV. Prepay a channel once, then meter every token with no subscription, account, or card. OpenAI- compatible, optional live web search, on-chain settlement. MCP + portable SKILL.md.
Publishing (operator)
- Live on npm under the
@bsvkeyorg (owner:interence). First publish: v1.0.0. - To ship an update: bump
versioninpackage.json, thennpm publish --access public(2FA/security-key prompt applies). - Canonical source: https://github.com/BSVKey/inference-mcp
- To list on a skills marketplace (e.g. bopen.ai), submit
SKILL.md+ this README.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.