> For the complete documentation index, see [llms.txt](https://docs.10102.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.10102.io/agents-and-builders.md). # Agents & Builders This page is for two audiences: developers building software (including AI agents) that needs to read or set up on-chain legacies, and autonomous agents that act on behalf of their users. It explains how to discover 10102 Computing Legacy as a registered agent, what you can call for free, and how metered inference works. The product is non-custodial and fully on-chain. Nothing here takes custody of funds, and no endpoint can move assets on its own: every contract write requires the user's wallet signature. ## Meet Guardian: the AI digital-legacy guide Guardian is the AI assistant inside 10102 Computing Legacy. It helps people set up their digital legacy and timelock contracts: it interviews a wallet owner about goals, helps them choose a matching contract type (Legacy or Timelock), and builds pre-filled setup links that the user completes by connecting a wallet and signing on-chain. * **Powered by Venice AI.** Guardian runs on Venice's privacy-preserving inference. Conversations are not stored and are not used to train models. The chat model is GLM 5.2. * **Free allowance for everyone.** A free inference allowance applies per caller, so anyone can use Guardian without bringing their own credits. * **A guide, not an advisor.** Guardian helps you configure legacy and timelock contracts. It does not give legal, tax, or financial advice. ## Discover us as an agent (ERC-8004) 10102 is registered under [ERC-8004](https://eips.ethereum.org/EIPS/eip-8004), the agent-discovery standard. Agent identities live in an Identity Registry implemented as an ERC-721 on Ethereum mainnet, so any agent or tool can resolve our identity on-chain. * **Agent name:** 10102 Guardian * **Agent id:** `34821` * **Agent registry:** `eip155:1:0x8004A169FB4a3325136EB29fA0ceB6D2e539a432` The identity is discoverable through standard ERC-8004 explorers such as [8004scan.io](https://8004scan.io). ### A2A agent card The agent capabilities are described by an A2A agent card (protocol version 0.3.0): ``` https://app.10102.io/.well-known/agent-card.json ``` The card advertises these skills: * **Suggest a legacy plan** (`recommend-legacy-plan`): interview a wallet owner about goals and help them choose the matching contract type. * **Build pre-filled setup URL** (`build-setup-url`): generate a deep link that pre-fills the web form with beneficiaries, durations, name, and type. The user finishes by connecting a wallet and signing on-chain. * **Query legacy status** (`query-legacy-status`): read LegacyDetail records (status, beneficiaries, triggers, Safe context) from the subgraph or on-chain router views. * **Query timelock status** (`query-timelock-status`): read Timelock entities (unlock time, lock type, recipient, token contents, withdrawal status). * **Verify on-chain deployment** (`verify-contract-deployment`): confirm canonical router addresses and read live contract state via Ethereum mainnet RPC. ## Model Context Protocol (MCP) Guardian exposes a read-only MCP server: ``` https://mcp.10102.io/mcp ``` (MCP protocol version `2025-06-18`.) Reads are free. The MCP server lets you look up legacies and timelocks and generate pre-filled setup links. There are no keys to manage and no write paths: the server returns indexed state and deep links, but contract creation still requires the user's wallet signature in the app or via a direct contract call. In practice that means an agent can: * Query indexed Legacy and Timelock state for an address. * Build a pre-filled setup URL that hands the user off to the web app to review and sign. ## Pay-per-call inference (x402) Guardian chat inference is available through an OpenAI-compatible endpoint (the venice-proxy): ``` https://venice-proxy-production.up.railway.app/v1/chat/completions ``` * It is **OpenAI-compatible**, so you can point an existing `/v1/chat/completions` client at it. * A **free tier** applies per caller, subject to a rate limit. * For unmetered, metered-by-you usage, bring your own Venice credits via [x402](https://www.x402.org) (the `X-Sign-In-With-X` header). The agent registration advertises `x402Support: true`, so you pay per call from your own wallet credits rather than relying on the shared free allowance. ## Build with us * **Source code:** the contracts and routers are open source at [github.com/10102-io/computing-sc](https://github.com/10102-io/computing-sc). * **Agent docs:** the canonical agent onboarding doc is published at [AGENTS.md](https://github.com/10102-io/computing/blob/main/AGENTS.md). * **Subgraph:** indexed Legacy and Timelock state is served from The Graph gateway: `https://gateway.thegraph.com/api/subgraphs/id/FXzZDbZxzdWaYGTPEcxN2AatbucRooQWfY67moam8oCd`. * **Non-custodial and on-chain:** every flow is operable directly from the Ethereum contracts. Reads are free, setup is a handoff to a wallet signature, and no part of this stack can move a user's assets without that signature.