# Put specialist tools within reach of your AI assistant

Connect a compatible AI assistant to Rutledge tools through Model Context Protocol. Give it specific work: find a source, look up a reference value or calculate a synthetic scenario that a professional reviews. Your organization controls which tools are on, and every call is recorded.

## Retirement and tax scenarios, side by side
`plan_simulate` runs an inline synthetic retirement scenario. It returns a projection, success measures and guardrail results across up to 200 simulated market paths in one call.

`tax_calculate` runs a single-year individual income-tax calculation in the Form 1040 shape. Results show the calculation version, federal and state outcomes, and a reason when an input cannot be calculated.

For example, ask your AI assistant to compare two fictional retirement-income choices, calculate the tax for each and explain the difference for your review. Check each tool's inputs; carrying assumptions from one calculation to the next is part of your review.

## Research with the source in view
`corpus_search` returns cited passages from the financial, tax and insurance reference collection. `corpus_grounded_search` adds commercial-insurance filters for jurisdiction, class and effective date.

These tools return the source text itself. Your AI assistant may use it to write an explanation; check that explanation against the cited passages. Legal professionals can use the collection for questions in financial, tax and insurance subjects.

## Reference tools for specialist questions
The tool set also includes:

- IRS Applicable Federal Rates and the section 7520 rate, plus economic data series.
- Public retirement-plan filings and fee benchmarks.
- Security lookup, price history and portfolio optimization.

The tools your assistant sees depend on your organization's enabled tool set. Check source coverage, dates and calculation assumptions every time.

## Use public information and synthetic scenarios
The MCP connection is for reference information and scenarios without client-identifying details. Use fictional inputs. Keep client names, tax identifiers, contact details, private documents and client-record references out of tool requests.

For approved integrations with practice records, use the separate [public API](/developers/api). Review your AI assistant's own handling of prompts and results alongside [Security & Control](/trust).

## Connect with organization authorization
The MCP endpoint uses Streamable HTTP at `/mcp` and OAuth sign-in. Setup needs a registered client and a signed-in user for one organization. We confirm the endpoint and sign-in setup with you; a website sign-in does not include MCP access.

Your organization decides whether MCP is on and which tools may be called. Tool limits and service permissions apply, and every call is recorded for audit. Use the server's `tools/list` descriptions and input schemas to build requests.

## Inspect a tool before using it
Once organization access is set up and your MCP client has connected, ask it to list the available tools:

```json
{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}
```

The response lists tool names, descriptions and input schemas.

## Call a reference lookup
If `gov_irs_rates` is enabled, this reference lookup takes no client inputs:

```json
{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"gov_irs_rates","arguments":{}}}
```

Its result contains the latest AFR and section 7520 reference records. Check their effective dates and source before using a rate. These are protocol messages inside an established MCP session; your client handles connection setup, sign-in and transport headers. Listing a tool does not skip the organization's permission checks.

## Your setup sequence
1. [Talk with us](/contact) about MCP setup: name your AI assistant, your organization and the first task you want to enable.
2. Confirm the endpoint, registered-client sign-in and enabled tool set with our team.
3. Connect, inspect `tools/list`, then run an enabled reference lookup. Review that output before moving to a fictional retirement or tax scenario.

## Plan a useful first integration
Tell us your AI assistant, the tasks you want it to do and the capabilities your team needs. Start with a reference lookup and a fictional scenario, then review the outputs, permitted usage and charges before wider use.

[Talk with us about MCP access](/contact) · [Pricing](/pricing) · [How AI is used](/ai)