Skip to main content

​
Glossary

TermDefinition
PassportUniversal identity primitive for models, compute nodes, tools, datasets, and agents. Each passport has a type, metadata, and capabilities. See Passports.
ReceiptCryptographic proof of an AI inference, containing input/output hashes, an Ed25519 signature, and a timestamp. See Receipts.
MMRMerkle Mountain Range — append-only data structure used to batch receipts into epochs with efficient inclusion proofs. See MMR.
EpochA batch of receipts whose MMR root is anchored to the Solana blockchain as a single transaction. See Epochs.
AnchoringThe process of committing an epoch’s MMR root hash to a Solana on-chain account, providing blockchain-level immutability.
Session SignerServer-side Ed25519 keypair that signs every receipt for cryptographic authenticity. See Session Signer.
Inclusion ProofMerkle proof demonstrating that a specific receipt is part of an anchored epoch’s MMR.
TrustGateOpenAI-compatible LLM proxy that routes inference requests via passport matching, with receipt generation. See TrustGate.
MCPGateMCP tool gateway with 88+ builtin servers, credential management, session budgets, and semantic discovery. See MCPGate.
MCPModel Context Protocol — open standard by Anthropic for connecting AI models to external tools and data sources.
PolicyRules for model matching — defines required capabilities, cost constraints, latency targets, and provider preferences.
Matching EngineThe @raijinlabs/passport service that selects the optimal model or compute node based on a policy.
Credential AdapterPluggable interface for resolving service credentials in MCPGate (EnvVar, Database, or Composite adapters).
Session BudgetPer-session resource cap in MCPGate — limits tool calls, cost, or duration with hard/soft enforcement.
Agent OrchestratorService managing the plan → accomplish → execute → validate lifecycle for autonomous AI agents.
PayoutRevenue split calculation between model provider, compute provider, and platform for each inference.
BYOKBring Your Own Key — use your own LLM provider API keys through Lucid, paying only infrastructure fees.
iGasInference gas — Solana fee unit consumed when anchoring inference receipts on-chain.
mGasMemory gas — Solana fee unit consumed when storing agent memory on-chain.
LiteLLMUnified LLM provider interface used by TrustGate to route requests to OpenAI, Anthropic, Google, Mistral, and open-source models.
OpenMeterUsage metering service that tracks API requests, tokens, receipts, and costs in real-time.
FlowSpecVisual workflow specification format (n8n-compatible) for defining multi-step AI pipelines.