Core Concepts

Understanding Agistry starts with four foundational components: Adapters, Registry, Capabilities, and Proofs. Together, these components form the foundation for standardized, decentralized interaction between autonomous agents and off-chain tools.

---

Adapters

Adapters are modular interface wrappers that define how a specific external tool: like an API, smart contract, or microservice, can be called. Think of them as programmable execution layers.

Each adapter includes:

• A capability schema (e.g. oracle_read, swap_tokens, fetch_data)

• Input and output types

• Optional authentication config

• Rate-limiting logic

• Endpoint abstraction (REST, gRPC, WebSocket, etc.)

• Optional cryptographic proof generation

Adapters are versioned, independently deployable, and can be published permissionlessly into the registry.

Adapters can wrap anything from a REST API to a Web3 contract call. Agents don’t care what’s under the hood — they just see the capability.

---

Registry

The Registry is the decentralized, index of all available adapters. It serves three core roles:

1. Discovery Layer Agents can query the registry to find adapters matching desired capabilities. For example, an agent might search for oracle_read or dex_swap tools with specific filters.

2. Metadata Store Each adapter includes metadata such as:

   • Creator/owner

   • Source repo or hash

   • Input/output schema

   • Permissions or licensing info

   • Proof protocol version

3. Version Management Adapters are immutable once published. Updates result in a new version, traceable via hash and registry index.

The registry ensures that every tool integration is public, auditable, and upgradable without breaking agent logic.

---

Capabilities

A capability is a high-level description of what an adapter does. This allows agents to reason about functionality without knowing the technical details of the underlying tool.

Examples of capabilities:

• price_fetcher/pyth/solana

• execute_swap/jupiter

• http_post/any

• read_state/smart_contract

Capabilities are defined in a semantic namespace format: domain/tool/chain, but can be extended arbitrarily.

Agents use these tags to filter and bind adapters into workflows.

Think of this as a universal "function catalog" for agent infrastructure.

---

Proofs

Proofs are cryptographic records of what was executed, when, and with what inputs/outputs.

Agistry optionally enables adapters to return signed proofs of execution. These include:

• Adapter hash

• Capability tag

• Input & output hash

• Timestamp

• Optional zkProof (in development)

• Agent signature or identity

Proofs allow:

• Replay of past executions

• Audit trails for regulators or stakeholders

• Trustless collaboration between agents

• Rollback logic or consensus formation in multi-agent systems

In security-critical or compliance-heavy environments, verifiable execution is non-negotiable. Agistry bakes this in as a native feature, not an afterthought.