The trust harness for AI, in every MCP client.
Connect Proofable once, and your agents can use your profile, proofs, permissions, and protected tools in Cursor, Claude, Codex, VS Code, ChatGPT, and any other MCP client.
Open Set up Proofable, choose your client, and finish sign-in. Then ask:
Show my Proofable profile and the proofs I can reuse.
For a specific requirement:
Check whether I already have the proof needed for this task. Reuse it if it qualifies; otherwise show me the next step.
Add https://mcp.proofable.me/mcp to any MCP client, or use a standard MCP configuration:
{
"mcpServers": {
"proofable": {
"type": "http",
"url": "https://mcp.proofable.me/mcp"
}
}
}The installer writes the same connection for supported clients:
npx -y @proofable/sdk setupClaude Code installs the server and the workflow skills together:
/plugin marketplace add proofable/mcp
/plugin install proofable-mcp@proofable
Keep one Proofable connection per client. A marketplace plugin plus a manual entry creates a duplicate.
| Outcome | Tools |
|---|---|
| Load the signed-in profile and current context | proofable_context |
| Find supported checks and their required inputs | proofable_verifiers_catalog |
| Check whether a qualifying proof already exists | proofable_proofs_check |
| Reuse a current proof or get the next verification step | proofable_verify_or_guide |
| Create or refresh a proof | proofable_verify |
| Prepare and submit a check needing another signer | proofable_verify_prepare, proofable_verify_submit |
| Find and read proofs | proofable_proofs_get |
| Update proof metadata or add feedback | proofable_proofs_update |
| Check, create, and load agent permissions | proofable_agent_link, proofable_agent_create, proofable_agent_mount |
Call proofable_context once after connecting. Proofable reuses current proofs before starting another verification and returns Passed, Action needed, or Blocked.
The MCP tool reference lists every tool with its inputs and result shapes.
Create an access key under Account → Access keys, store it as PROOFABLE_ACCESS_KEY, and send it as a Bearer token. Keep access keys out of source code, browser bundles, and chat.
{
"mcpServers": {
"proofable": {
"type": "http",
"url": "https://mcp.proofable.me/mcp",
"headers": {
"Authorization": "Bearer ${PROOFABLE_ACCESS_KEY}"
}
}
}
}- Prompt injection test: hides AgentDojo attacks in a note your agent reads, runs the task under a scoped and a broad permission, and grades every run from its signed receipt.
- Guarded agent action: makes one permission decision before host code runs a tool.
Use the Proofable SDK for application code, verification gates, and the CLI. This package carries the MCP registry manifest, the host plugins, and the workflow skills; the server runs at https://mcp.proofable.me/mcp.
Apache-2.0. Proofable is published by NEUS Network, Inc.