Kit
io.github.torevaPublished by GitHub account torevav0.2.0Updated Oct 24, 2026
Policy-checked Solana wallet actions with user-set limits, unsigned transactions, and receipts.
Overview
Lets an assistant read Solana wallets, simulate and prepare unsigned transactions under user-set limits, and fetch receipts.
- What you can do
- Kit is a local stdio MCP server for Toreva governed objects. Its tools create an agent wallet with hard limits, read or scan a main or agent wallet, simulate an action before anything is prepared, prepare an unsigned transaction after simulation, explain an action or receipt, fetch a single receipt, and refuse an unsafe action with a refusal receipt. Reads and writes go through the Toreva gateway relay, and accepted, pending, or refused outcomes return a receipt.
- When to use it
- Use it when an assistant should inspect Solana wallets and prepare policy-checked, unsigned transactions that the user still signs, with receipts as proof of each outcome. It suits users who want simulation and refusal paths before any transaction is prepared.
- Requirements
- Runs as a local process over stdio, installed from the npm package @toreva/mcp (Node.js). Needs a Toreva login token or an explicit RELAY_AUTH_TOKEN from your own secret store, and network access to the Toreva gateway relay. Supported client config targets are claude-desktop, cursor, and openclaw.
Installation
In SourceWeft
- Open Kit in the dashboard and add it to a workspace.
- Enable the server for the chats that should use its tools.
Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.
Other MCP clients
Follow the launch instructions in the repository.
README
toreva kit
Connector primitives for Toreva governed objects. Bring your own compute, use the gateway relay, and require receipts for every outcome.
Kit provides SDK, CLI, and MCP connectors for developers connecting their own compute to Toreva governed objects. The setup paths below are developer integrations.
The Day-1 Mac helper scope is Pair and Unpair only, for every user. It must request no Screen Recording or Accessibility permissions and take no jobs. Developer connector setup is separate from helper pairing.
The front door is the object store. MCP clients, SDK callers, CLI users, and customer-owned agent fleets all read and write the same Toreva objects through the gateway relay, and every accepted, pending, or refused outcome returns a receipt.
GitHub can prove which organization and admin are acting. It is not the transport for Toreva work messages. Compute is bound per scope, and the live paths put model execution on the developer's own account.
No money moves through this connector.
The Three Doors
Door 1: MCP
For developer integrations, start here. Your AI client connects the Toreva MCP server, your own client subscription does the work, and Toreva receives only the object reads and writes you authorize.
Restart the client, then ask it about your Toreva objects or receipts. Supported
client config targets are claude-desktop, cursor, and openclaw.
Door 2: Customer-Owned Fleet
Use this when you need unattended or scheduled agents. A GitHub organization
admin runs agents in their own org, on their own runners, with their own model
credentials. The agents call Toreva through MCP, SDK, or CLI and land answers
back on Toreva objects with receipt_id proof.
Start here: GitHub Org Agent Fleet.
Door 3: Hosted Compute
Hosted compute means Toreva runs the agents and carries the marginal model cost. This door is deferred and is not documented as available. Use MCP or a customer-owned fleet today.
How Answers Land
The object store is the receiving system. A work surface prepares a relay
request, sends it to https://gateway.toreva.com/relay, and treats the result
as accepted only when the gateway returns a receipt_id.
For the shared install, invoke, and receipt shapes, see Work-Surface Connector Primitives.
MCP Tools
Install
The local package runs over stdio. Use either the token written by
toreva login or an explicit RELAY_AUTH_TOKEN from your own secret store.
For runner-owned environments:
For local development:
Package
@toreva/mcp is the Rung-1 MCP package. Version 0.2.0 is stdio-local only.
The public metadata is server.json and packages/mcp/package.json.
Packages
Verify
Regulatory notice
This software provides tooling for interacting with the Toreva policy service. It does not provide financial advice, investment advice, trading advice, or any other form of advice. Use of this software does not create a fiduciary relationship, advisory relationship, or any other professional relationship between you and Toreva Pty Ltd.
Toreva Pty Ltd is not responsible for modifications made to this software by third parties, including modifications that alter or remove compliance language, disclaimers, or risk warnings. If you use a modified version of this software, you do so at your own risk and are responsible for ensuring your use complies with applicable law.
License
MIT — see LICENSE
Source: README.md at commit 793dc5e
Tools
0Version history
1- v0.2.0LatestSep 16, 2026
Score
40% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 days25 pts
Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.
- Auth declared10 pts
Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts
Declare a license in the repository or the registry entry.