io.github.prismnetwork-tech/mcp

Read the output of a batch lease by lease_id, once its node has reported. Use it to recover a result after prism_batch_run timed out.

Fund a lease that runs one command with no interactive access at all: the node executes it and reports the signed output. Matches only suppliers at trust class 'isolated' or above, so it can find no supplier when none is online; prefer prism_lease_and_run for broad availability. Output is capped at 64 KiB per stream.

Show the spending limits this server enforces and what it has already spent in the last 24 hours, with the recent charges. Needs no wallet. Check this before a long job; a lease refused for budget says the same numbers.

Buy one LLM generation that runs inside a GPU TEE, with the message contents encrypted end to end to a key the enclave's own attestation commits to, so Prism's relay in between carries ciphertext and cannot read the prompt or the answer. Costs a little more than prism_infer. Returns the answer, the cost, and a receipt id the workload signed over the exact bytes of the exchange; pass that id to prism_verify_attestation to check the whole chain. Use it for anything the operator of an ordinary endpoint should not be able to read.

Release a lease. Access closes and billing stops here: settlement charges the seconds the lease was open and returns the rest of the deposit. A lease nobody releases bills until its window ends.

Buy one LLM generation from Prism's managed inference endpoint. Pays the quoted USDG price from this wallet (about 0.01 USDG), waits through a cold start when no box is warm (up to a few minutes), and returns the generation with token usage. Cheaper and simpler than leasing when all you need is a completion.

Buy many LLM generations from Prism's managed inference endpoint in one paid call. Every prompt runs whole on a rented GPU, spread across every GPU the endpoint holds, so a list of prompts finishes far sooner than the same prompts sent one at a time. Costs the single-generation price times the number of prompts. Returns every answer in order plus a Merkle receipt naming the leases that did the work. Use it for evals, dataset passes, rollouts, or anything with more than a handful of independent prompts.

Lease a GPU and keep it running. Returns a lease_id and SSH access. Use prism_run to execute commands and prism_end_lease when done.

Lease a GPU, run one shell command on it, and return the output. The lease stays alive (use prism_run for more commands, prism_end_lease to release). Prefer this for a single command; use prism_lease when you'll run several.

List this wallet's leases on Prism Network with their current state.

List GPUs currently available to lease on Prism Network, with model, VRAM, price per second in USDG, and trust class. Trust class runs open < isolated < attested < confidential; on an 'open' supplier the host operator can read anything the workload touches. Keep secrets and credentials in prism_vault_store rather than on the box, and raise min_trust when the workload itself must not be readable.

Current GPU pricing on Prism Network by model: sourced low/median/high and settled mean, in USDG per hour. Needs no wallet. Use it to estimate what an analysis job will cost before leasing.

Recent settled lease receipts from the public proof feed: GPU model, runtime, what was charged and refunded, and the settlement transaction hash on Robinhood Chain. Needs no wallet. Every Prism lease ends in one of these.

Run a shell command on a GPU you already leased with prism_lease.

Permanently delete a vault item. The ciphertext is removed and the value cannot be recovered.

List the agent's sealed vault items: item_id, label, version and trust floor. Values are not returned and are not readable by Prism.

Decrypt and return one vault item, in this process, using the wallet-derived key. The plaintext exists only here; do not echo it into a leased workspace, a log, or a message.

Authorize a vault item into a lease you hold and return its plaintext for use there. Refused when the lease's trust class is below the item's trust floor, which is what stops a secret reaching a host that can read it. Every allowed release is recorded against the account.

Store private data (a card, an identity document, an API credential) encrypted under a key derived from this agent's wallet on this machine. Prism receives ciphertext only and cannot read it. Use this instead of writing a secret into a workspace or a file. Returns an item_id; the value is not recoverable without the wallet.

Check a confidential generation against its signed receipt and the hardware behind it: the TDX quote verifies to Intel's root and commits to the key set that signed the receipt, the boot log replays to the measurement in that quote, the receipt covers the exact bytes of this call, the upstream that ran the model was itself verified, and the GPU is attested by NVIDIA. Returns every check with its result, including the ones that cannot be established today. Needs no wallet.

Show the agent's wallet address and on-chain balances (USDG and ETH for gas) on Robinhood Chain. Check this before leasing to confirm the wallet can pay.