AI Agent Signal Source

Connect your AI agent to controlled trading operations.

Vorda-hosted MCP starts with trade proposals and is designed to expand into scoped account visibility and controlled execution tools.

Today the agent supplies trading intent through vorda.propose_action. As additional tools ship, the same connection can expose positions, execution history, previews, and approved actions without giving the model broker credentials or authority outside a running Automation.

Create AI Agent source Choose a client

Agent proposalMCP

Calls vorda.propose_action

Vorda Signal Source

Proposal acceptedAwaiting Automation

Source identity verified
Intent normalized
No trade during source test

Choose your AI client

Create one AI Agent Signal Source, then use the connection method generated for your client.

ChatGPTView setup ClaudeView setup Claude CodeView setup CursorView setup VS CodeView setup PyCharmView setup AntigravityView setup Hermes AgentView setup OpenClawView setup OdysseusAIView setup Codex or custom MCP clientView setup Another MCP clientAsk us

Create the source in Vorda first

Vorda generates source-specific connection details. There is no universal public endpoint or shared workspace token to copy from this website.

Create AI Agent source

Choose AI Agent on the New Signal Source page.
Name the source and select ChatGPT, Claude, Codex, or Custom.
Copy the generated MCP URL and complete the client-specific authentication flow.
Send a test proposal and confirm the normalized payload. Testing cannot execute a trade.
Create an Automation only when you want to grant execution authority.

MCP server: Generated per Signal Source
Tool: vorda.propose_action
Test mode: Validation only, never execution
Execution authority: A separate running Automation

Client setup

ChatGPT

Connect ChatGPT to an AI Agent Signal Source created in Vorda.

Create an AI Agent Signal Source in Vorda and choose this client.
Use the generated MCP connection details and the OAuth flow shown by Vorda.
Run the safe source test. It validates the proposal and never places a trade.
Link the source to an Automation with an exact Rules version and compatible broker account when you are ready.

Keep private: Keep the source-specific connection URL and credential private. Rotate them in Vorda if exposed.

Agent instruction

Give the client a narrow instruction that matches Vorda's authority boundary.

First discover which Vorda MCP tools are available in this workspace.

Today, use vorda.propose_action only when I explicitly ask you to propose a trade. Include the requested trading intent, such as symbol, side, quantity, order type, price constraints, and a unique client_reference.

When Vorda adds account visibility, execution history, preview, or controlled execution tools, use only tools that are actually discovered and authorized. Present the execution plan before any action that could change an order or position. Never claim that a missing tool ran or that a proposal executed. Never choose a broker, change Trading Rules, expose credentials, or bypass Automation readiness.

Client setup

Claude

Connect Claude to an AI Agent Signal Source created in Vorda.

Create an AI Agent Signal Source in Vorda and choose this client.
Use the generated MCP connection details and the OAuth flow shown by Vorda.
Run the safe source test. It validates the proposal and never places a trade.
Link the source to an Automation with an exact Rules version and compatible broker account when you are ready.

Keep private: Keep the source-specific connection URL and credential private. Rotate them in Vorda if exposed.

Agent instruction

Give the client a narrow instruction that matches Vorda's authority boundary.

First discover which Vorda MCP tools are available in this workspace.

Today, use vorda.propose_action only when I explicitly ask you to propose a trade. Include the requested trading intent, such as symbol, side, quantity, order type, price constraints, and a unique client_reference.

When Vorda adds account visibility, execution history, preview, or controlled execution tools, use only tools that are actually discovered and authorized. Present the execution plan before any action that could change an order or position. Never claim that a missing tool ran or that a proposal executed. Never choose a broker, change Trading Rules, expose credentials, or bypass Automation readiness.

Client setup

Claude Code

Connect Claude Code to an AI Agent Signal Source created in Vorda.

Create an AI Agent Signal Source in Vorda and choose this client.
Use the generated MCP connection details and the Custom client connection generated by Vorda.
Run the safe source test. It validates the proposal and never places a trade.
Link the source to an Automation with an exact Rules version and compatible broker account when you are ready.

Keep private: Keep the source-specific connection URL and credential private. Rotate them in Vorda if exposed.

Agent instruction

Give the client a narrow instruction that matches Vorda's authority boundary.

First discover which Vorda MCP tools are available in this workspace.

Today, use vorda.propose_action only when I explicitly ask you to propose a trade. Include the requested trading intent, such as symbol, side, quantity, order type, price constraints, and a unique client_reference.

When Vorda adds account visibility, execution history, preview, or controlled execution tools, use only tools that are actually discovered and authorized. Present the execution plan before any action that could change an order or position. Never claim that a missing tool ran or that a proposal executed. Never choose a broker, change Trading Rules, expose credentials, or bypass Automation readiness.

Client setup

Cursor

Connect Cursor to an AI Agent Signal Source created in Vorda.

Create an AI Agent Signal Source in Vorda and choose this client.
Use the generated MCP connection details and the Custom client connection generated by Vorda.
Run the safe source test. It validates the proposal and never places a trade.
Link the source to an Automation with an exact Rules version and compatible broker account when you are ready.

Keep private: Keep the source-specific connection URL and credential private. Rotate them in Vorda if exposed.

Agent instruction

Give the client a narrow instruction that matches Vorda's authority boundary.

First discover which Vorda MCP tools are available in this workspace.

Today, use vorda.propose_action only when I explicitly ask you to propose a trade. Include the requested trading intent, such as symbol, side, quantity, order type, price constraints, and a unique client_reference.

When Vorda adds account visibility, execution history, preview, or controlled execution tools, use only tools that are actually discovered and authorized. Present the execution plan before any action that could change an order or position. Never claim that a missing tool ran or that a proposal executed. Never choose a broker, change Trading Rules, expose credentials, or bypass Automation readiness.

Client setup

VS Code

Connect VS Code to an AI Agent Signal Source created in Vorda.

Create an AI Agent Signal Source in Vorda and choose this client.
Use the generated MCP connection details and the Custom client connection generated by Vorda.
Run the safe source test. It validates the proposal and never places a trade.
Link the source to an Automation with an exact Rules version and compatible broker account when you are ready.

Keep private: Keep the source-specific connection URL and credential private. Rotate them in Vorda if exposed.

Agent instruction

Give the client a narrow instruction that matches Vorda's authority boundary.

First discover which Vorda MCP tools are available in this workspace.

Today, use vorda.propose_action only when I explicitly ask you to propose a trade. Include the requested trading intent, such as symbol, side, quantity, order type, price constraints, and a unique client_reference.

When Vorda adds account visibility, execution history, preview, or controlled execution tools, use only tools that are actually discovered and authorized. Present the execution plan before any action that could change an order or position. Never claim that a missing tool ran or that a proposal executed. Never choose a broker, change Trading Rules, expose credentials, or bypass Automation readiness.

Client setup

PyCharm

Connect PyCharm to an AI Agent Signal Source created in Vorda.

Create an AI Agent Signal Source in Vorda and choose this client.
Use the generated MCP connection details and the Custom client connection generated by Vorda.
Run the safe source test. It validates the proposal and never places a trade.
Link the source to an Automation with an exact Rules version and compatible broker account when you are ready.

Keep private: Keep the source-specific connection URL and credential private. Rotate them in Vorda if exposed.

Agent instruction

Give the client a narrow instruction that matches Vorda's authority boundary.

First discover which Vorda MCP tools are available in this workspace.

Today, use vorda.propose_action only when I explicitly ask you to propose a trade. Include the requested trading intent, such as symbol, side, quantity, order type, price constraints, and a unique client_reference.

When Vorda adds account visibility, execution history, preview, or controlled execution tools, use only tools that are actually discovered and authorized. Present the execution plan before any action that could change an order or position. Never claim that a missing tool ran or that a proposal executed. Never choose a broker, change Trading Rules, expose credentials, or bypass Automation readiness.

Client setup

Antigravity

Connect Antigravity to an AI Agent Signal Source created in Vorda.

Create an AI Agent Signal Source in Vorda and choose this client.
Use the generated MCP connection details and the Custom client connection generated by Vorda.
Run the safe source test. It validates the proposal and never places a trade.
Link the source to an Automation with an exact Rules version and compatible broker account when you are ready.

Keep private: Keep the source-specific connection URL and credential private. Rotate them in Vorda if exposed.

Agent instruction

Give the client a narrow instruction that matches Vorda's authority boundary.

First discover which Vorda MCP tools are available in this workspace.

Today, use vorda.propose_action only when I explicitly ask you to propose a trade. Include the requested trading intent, such as symbol, side, quantity, order type, price constraints, and a unique client_reference.

When Vorda adds account visibility, execution history, preview, or controlled execution tools, use only tools that are actually discovered and authorized. Present the execution plan before any action that could change an order or position. Never claim that a missing tool ran or that a proposal executed. Never choose a broker, change Trading Rules, expose credentials, or bypass Automation readiness.

Client setup

Hermes Agent

Connect Hermes Agent to an AI Agent Signal Source created in Vorda.

Create an AI Agent Signal Source in Vorda and choose this client.
Use the generated MCP connection details and the Custom client connection generated by Vorda.
Run the safe source test. It validates the proposal and never places a trade.
Link the source to an Automation with an exact Rules version and compatible broker account when you are ready.

Keep private: Keep the source-specific connection URL and credential private. Rotate them in Vorda if exposed.

Agent instruction

Give the client a narrow instruction that matches Vorda's authority boundary.

First discover which Vorda MCP tools are available in this workspace.

Today, use vorda.propose_action only when I explicitly ask you to propose a trade. Include the requested trading intent, such as symbol, side, quantity, order type, price constraints, and a unique client_reference.

When Vorda adds account visibility, execution history, preview, or controlled execution tools, use only tools that are actually discovered and authorized. Present the execution plan before any action that could change an order or position. Never claim that a missing tool ran or that a proposal executed. Never choose a broker, change Trading Rules, expose credentials, or bypass Automation readiness.

Client setup

OpenClaw

Connect OpenClaw to an AI Agent Signal Source created in Vorda.

Create an AI Agent Signal Source in Vorda and choose this client.
Use the generated MCP connection details and the Custom client connection generated by Vorda.
Run the safe source test. It validates the proposal and never places a trade.
Link the source to an Automation with an exact Rules version and compatible broker account when you are ready.

Keep private: Keep the source-specific connection URL and credential private. Rotate them in Vorda if exposed.

Agent instruction

Give the client a narrow instruction that matches Vorda's authority boundary.

First discover which Vorda MCP tools are available in this workspace.

Today, use vorda.propose_action only when I explicitly ask you to propose a trade. Include the requested trading intent, such as symbol, side, quantity, order type, price constraints, and a unique client_reference.

When Vorda adds account visibility, execution history, preview, or controlled execution tools, use only tools that are actually discovered and authorized. Present the execution plan before any action that could change an order or position. Never claim that a missing tool ran or that a proposal executed. Never choose a broker, change Trading Rules, expose credentials, or bypass Automation readiness.

Client setup

OdysseusAI

Connect OdysseusAI to an AI Agent Signal Source created in Vorda.

Create an AI Agent Signal Source in Vorda and choose this client.
Use the generated MCP connection details and the Custom client connection generated by Vorda.
Run the safe source test. It validates the proposal and never places a trade.
Link the source to an Automation with an exact Rules version and compatible broker account when you are ready.

Keep private: Keep the source-specific connection URL and credential private. Rotate them in Vorda if exposed.

Agent instruction

Give the client a narrow instruction that matches Vorda's authority boundary.

First discover which Vorda MCP tools are available in this workspace.

Today, use vorda.propose_action only when I explicitly ask you to propose a trade. Include the requested trading intent, such as symbol, side, quantity, order type, price constraints, and a unique client_reference.

When Vorda adds account visibility, execution history, preview, or controlled execution tools, use only tools that are actually discovered and authorized. Present the execution plan before any action that could change an order or position. Never claim that a missing tool ran or that a proposal executed. Never choose a broker, change Trading Rules, expose credentials, or bypass Automation readiness.

Client setup

Codex or a custom MCP client

Connect Codex or a custom MCP client to an AI Agent Signal Source created in Vorda.

Create an AI Agent Signal Source in Vorda and choose this client.
Use the generated MCP connection details and the generated bearer credential.
Run the safe source test. It validates the proposal and never places a trade.
Link the source to an Automation with an exact Rules version and compatible broker account when you are ready.

Keep private: Keep the source-specific connection URL and credential private. Rotate them in Vorda if exposed.

Agent instruction

Give the client a narrow instruction that matches Vorda's authority boundary.

First discover which Vorda MCP tools are available in this workspace.

Today, use vorda.propose_action only when I explicitly ask you to propose a trade. Include the requested trading intent, such as symbol, side, quantity, order type, price constraints, and a unique client_reference.

When Vorda adds account visibility, execution history, preview, or controlled execution tools, use only tools that are actually discovered and authorized. Present the execution plan before any action that could change an order or position. Never claim that a missing tool ran or that a proposal executed. Never choose a broker, change Trading Rules, expose credentials, or bypass Automation readiness.

A proposal is not an execution

MCP is the transport for the AI Agent Signal Source and the future agent-operations surface. New tools must preserve the same Automation, permission, Rules, and proof boundaries.

01
Current availability starts with vorda.propose_action. Position, balance, history, preview, and execution tools remain unavailable until they are explicitly shipped and discovered by the client.
02
Future account tools stay scoped. The agent still cannot select an unauthorized broker, change Trading Rules, expose credentials, or start an Automation.
03
A source test validates and displays a sanitized proposal without creating an order or broker request.
04
Only a compatible, ready, running Automation can evaluate and route a live proposal.
05
Every result remains attributable to its source, Rules version, target, and Automation.

Current and planned agent capabilities

The client should discover tools at runtime. Proposal requests work first; broader visibility and controlled actions become available only when Vorda exposes and authorizes those tools.

Current proposal workflow

Propose a market buy for 10 AAPL shares with client reference test-001
Propose a BTCUSDT sell with a limit price and explain the fields before sending
Show me the normalized proposal returned by Vorda

Planned scoped tools

Inspect authorized balances and current positions
Review recent executions, failures, and reconciliation state
Preview an intended order and its expected effect
Submit, cancel, close, or resize only through explicit workspace policy and Automation controls

FAQ

AI Agent and MCP questions

Are AI Agent and MCP separate Signal Sources?

No. AI Agent is the Signal Source type. MCP is the hosted transport used to deliver proposals from a supported AI client.

Can the agent place a trade by itself?

No. It calls vorda.propose_action. Execution requires a separate ready and running Automation with an exact Rules version and compatible broker accounts.

Will agents be able to inspect positions and manage approved trades?

That is the planned direction. Vorda will add scoped visibility, history, preview, and controlled action tools. The client must discover each tool at runtime, and every action remains subject to workspace permissions, Automation readiness, Trading Rules, and Execution Proof.

Can I test the source before creating an Automation?

Yes. The safe test validates the connection and normalized proposal, and it can never create an order or broker request.

Does the AI client receive my broker keys?

No. Broker credentials remain inside Vorda and are never included in the MCP connection or agent prompt.