{
  "markdown": "<!-- BEGIN TOREVA GITHUB LOGO -->\n<p align=\"center\">\n  <a href=\"https://github.com/toreva\">\n    <img src=\"https://raw.githubusercontent.com/toreva/.github/main/assets/toreva-logo-cyan-dark.png\" alt=\"Toreva\" width=\"96\">\n  </a>\n</p>\n<!-- END TOREVA GITHUB LOGO -->\n\n# toreva kit\n\nKit is the public client layer for connecting user-owned compute to Toreva\ngoverned objects.\n\nThe front door is the object store. MCP clients, SDK callers, CLI users, and\ncustomer-owned agent fleets all read and write the same Toreva objects through\nthe gateway relay, and every accepted, pending, or refused outcome returns a\nreceipt.\n\nGitHub can prove which organization and admin are acting. It is not the\ntransport for Toreva work messages. Compute is bound per scope, and the live\npaths put model execution on the developer's own account.\n\nNo money moves through this connector.\n\n## The Three Doors\n\n### Door 1: MCP\n\nUse this first. Your AI client connects the Toreva MCP server, your own client\nsubscription does the work, and Toreva receives only the object reads and writes\nyou authorize.\n\n```bash\nnpx toreva init --client=claude-desktop\nnpx toreva login\nnpx toreva doctor\n```\n\nRestart the client, then ask it about your Toreva objects or receipts. Supported\nclient config targets are `claude-desktop`, `cursor`, and `openclaw`.\n\n### Door 2: Customer-Owned Fleet\n\nUse this when you need unattended or scheduled agents. A GitHub organization\nadmin runs agents in their own org, on their own runners, with their own model\ncredentials. The agents call Toreva through MCP, SDK, or CLI and land answers\nback on Toreva objects with `receipt_id` proof.\n\nStart here: [GitHub Org Agent Fleet](./docs/github-org-agent-fleet.md).\n\n### Door 3: Hosted Compute\n\nHosted compute means Toreva runs the agents and carries the marginal model\ncost. This door is deferred and is not documented as available. Use MCP or a\ncustomer-owned fleet today.\n\n## How Answers Land\n\nThe object store is the receiving system. A work surface prepares a relay\nrequest, sends it to `https://gateway.toreva.com/relay`, and treats the result\nas accepted only when the gateway returns a `receipt_id`.\n\nFor the shared install, invoke, and receipt shapes, see\n[Work-Surface Connector Primitives](./docs/work-surface-connector-primitives.md).\n\n## MCP Tools\n\n| Tool | What it does |\n| --- | --- |\n| `toreva_establish_agent_wallet` | Create the agent wallet with hard limits |\n| `toreva_read_or_scan` | Read or scan a main wallet or agent wallet |\n| `toreva_simulate_action` | Test an action before anything is prepared |\n| `toreva_prepare_unsigned_transaction` | Prepare an unsigned transaction after simulation |\n| `toreva_explain_action` | Explain an action or receipt |\n| `toreva_get_receipt` | Fetch one receipt |\n| `toreva_refuse_action` | Refuse an unsafe action and return a refusal receipt |\n\n## Install\n\nThe local package runs over stdio. Use either the token written by\n`toreva login` or an explicit `RELAY_AUTH_TOKEN` from your own secret store.\n\n```bash\nnpm install -g @toreva/mcp\ntoreva login\ntoreva-mcp\n```\n\nFor runner-owned environments:\n\n```bash\nRELAY_AUTH_TOKEN=your_token toreva-mcp\n```\n\nFor local development:\n\n```bash\npnpm install\npnpm --filter @toreva/mcp build\nRELAY_AUTH_TOKEN=your_token node packages/mcp/dist/index.js\n```\n\n## Package\n\n`@toreva/mcp` is the Rung-1 MCP package. Version `0.2.0` is stdio-local only.\n\nThe public metadata is `server.json` and `packages/mcp/package.json`.\n\n## Packages\n\n| Package | What |\n| --- | --- |\n| `@toreva/mcp` | MCP server for agent integration |\n| `@toreva/types` | Shared schemas and types |\n| `@toreva/sdk` | TypeScript client library |\n| `@toreva/cli` | Command-line interface |\n\n## Verify\n\n```bash\npnpm test\npnpm --filter @toreva/mcp build\n```\n\n## Regulatory notice\n\nThis software provides tooling for interacting with the Toreva policy service.\nIt does not provide financial advice, investment advice, trading advice, or any\nother form of advice. Use of this software does not create a fiduciary\nrelationship, advisory relationship, or any other professional relationship\nbetween you and Toreva Pty Ltd.\n\nToreva Pty Ltd is not responsible for modifications made to this software by\nthird parties, including modifications that alter or remove compliance language,\ndisclaimers, or risk warnings. If you use a modified version of this software,\nyou do so at your own risk and are responsible for ensuring your use complies\nwith applicable law.\n\n## License\n\nMIT — see [LICENSE](./LICENSE)\n",
  "bytes": 4495,
  "sha": "3a32876e0344108ca66136fd1bc8e1a4d2bbae0ac08b946c90d5f1ff703f68f6",
  "repo_slug": "toreva/kit",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/mcp_io_github_toreva_kit_934dfbe4/readme"
}