What Corridor MCP Can Do

Corridor MCP gives agents access to information in Corridor. All requests use the user’s Corridor permissions. If you cannot access something in the platform, MCP cannot access it for you either. See Input and Output Guardrails.

Some of the things Corridor MCP can do are:

What it can do

Find objects

Search registered objects by name, alias, owner, status, group, or tags. For example:

• “Do we already have a feature for revolving utilisation?”
• “Show me approved models in the Auto Lending group.”
• “Find models that are still in pending approval?”

Search is based on registered metadata. It does not search through the contents of object definitions or code.

Understand objects and dependencies

Ask what an object does, what it uses, or what depends on it. For example:

• “What inputs are used in this model?”
• “Show me all features developed on this columns from this table?”

You can also ask about an object’s definition, inputs, and other metadata. For models, this includes custom fields, which can be used to answer questions such as “Which models are due for review next month?”

Analyse jobs and reports

Once a job has run, ask about what it produced. For example:

• “What was the default rate by segment in this run?”
• “Which reports did this job generate?”

Job output can be queried directly, so counts and metrics come back as exact values rather than estimates from a summary. Reports and their attachments can be read the same way.

Diagnose failures

When a run fails, the error logs for that job can be read and explained. For example:

• “Why did this simulation fail?”
• “What went wrong in the last run of this policy?”

Write object definitions

Ask for help writing the code for an object definition. Corridor MCP supplies the context needed to get the format right, such as an existing report definition to work from.

Generated code is a starting point. It still goes through the same testing and approval path as anything written by hand.

Search documentation

Ask questions about the Corridor platform and get answers based on the product documentation. Responses include a link to the relevant documentation page so you can check the source.

The main areas covered are:

Area	What it covers
Objects	Search and object metadata
Policies	Strategies, segments, rules, and outcomes
Jobs	Job details and output data
Reports	Reports, metrics, and report attachments
Lineage	Upstream and downstream dependencies
Documentation	Documentation search and content
Attachments	Stored files and job output