RiskRover MCP for VS Code

Draft and refine risk-based test plans in RiskRover and assess automated test coverage against product risks using VS Code AI agents.

Work with your agent on the Product Breakdown Structure (PBS), Product Risk Analysis (PRA), Test Strategy, Effort Estimation and Budget. Review each proposal before approving changes to your RiskRover project.

Use the same project risks to assess automated tests in your open workspace folders and identify coverage gaps. The assessment should distinguish what the tests cover from what available execution evidence shows.

The extension bundles the local MCP server, so this setup does not require a separate Node.js installation. A verified email, an active Builder or Pro subscription, and access to the project are required. Builder provides the same MCP planning capabilities within the Free plan’s limits; execution, exports, collaboration and general API access require Pro.

Planning steps supported

Set up your project in RiskRover first, then work with your AI agent through steps 3–7. Read-only access supports drafting; saving reviewed changes requires write access and your explicit approval. The MCP cannot change the overall project budget amount or currency, or delete or reorder roles.

Sign in and select a project

1. Open and trust a local workspace in VS Code.

2. Run RiskRover MCP: Sign In and choose GitHub Copilot or Codex.

3. Sign in to RiskRover in the browser, completing two-factor authentication if enabled.

4. Search for and select the target project, choose read-only or read-and-reviewed-write access, and approve the connection.

5. Wait for the browser to confirm setup is complete, then close it and return to VS Code.

6. Start a new agent session for the selected project. For Copilot, first restart RiskRover through MCP: List Servers. For Codex, trust the workspace if prompted; if the connection has not updated, run Developer: Reload Window and start a new session. Keep VS Code open.

RiskRover MCP: Select Project repeats this flow for the current workspace and selected host. The project is a default; explicit access to other authorized projects still works. Connections for different workspaces and hosts are independent.

Start a new chat after reconnecting or switching projects. An old MCP process will report that the connection changed; reload the window if Codex keeps that process running. Test IDE Connection checks the saved connection, not an existing chat's configuration.

The Codex path writes a secret-free .codex/config.toml in each local workspace folder, preserving unrelated settings. It uses a local bridge while VS Code remains open; Codex does not read SecretStorage itself. Invalid TOML or failed verification leaves the previous connection in place. Machine-specific paths should stay out of shared configuration.

Use RiskRover MCP: Test IDE Connection to verify access and RiskRover MCP: Sign Out to disconnect the selected host in this workspace. Browser grants expire after 90 days and can also be revoked in RiskRover Security > Connected IDEs.

MCP: Browse MCP Servers is discovery. Use MCP: List Servers for configured/runtime status.

Manual token alternative

For a new manual setup, create an MCP operations only token under Manage workspace > Configuration > API TOKENS in RiskRover. Existing supported Pro API tokens continue to work. Read-only scope supports assessment and drafting; saving reviewed changes requires read-and-write scope. Each person uses their own token.

For Copilot, optionally set riskroverMcp.projectId, run RiskRover MCP: Set API Token, then RiskRover MCP: Test Connection. A rejected token preserves the previous one. Saving after a transport-only verification failure requires explicit confirmation.

For Codex, RiskRover MCP: Copy Codex Configuration copies a safe placeholder and a stable bundled-server path for private ~/.codex/config.toml. Replace the placeholder yourself. Never commit tokens. Keep only one RiskRover registration in each host, restart it, and start a new agent session after configuration changes.

If you prefer the portable npm package, have Node.js 20 or newer installed and use this private Codex configuration instead:

[mcp_servers.riskrover]
command = "npx"
args = ["-y", "riskrover-mcp-server@latest"]
startup_timeout_sec = 60

[mcp_servers.riskrover.env]
RISKROVER_API_TOKEN = "PASTE_YOUR_PERSONAL_RISKROVER_API_TOKEN_HERE"
RISKROVER_PROJECT_ID = "12345"

Do not configure both the bundled path and the npm path under different names, because that exposes duplicate RiskRover tools.

Ask in normal language

Draft and refine a test plan:

Use RiskRover MCP to help me draft a risk-based test plan for project 12345.
Read the project context and test levels, then work through the PBS and
quality assignments, product risk analysis, test strategy, effort estimation,
and budget roles, rates and allocations. Work one step at a time, ask for
missing business inputs, and show each proposal for review. Do not submit
changes until I explicitly approve them.

Draft a product breakdown:

Use RiskRover MCP to draft a complete business-readable PBS for RiskRover
project 12345. Present it for review and do not import it yet.

Assess automated test coverage:

Use RiskRover MCP to assess automated test coverage in the workspace folders
I have made available against the product risks in project 12345. Keep it
read-only, identify coverage gaps, distinguish defined tests from available
execution evidence, and state which folders you inspected and any limitations.

RiskRover MCP supplies project and risk context; your AI client inspects the evidence it is permitted to access. The MCP server itself does not read repository files or run tests.

Refine individual planning steps

You can also ask your agent to refine a specific part of the plan:

• Set PRA likelihood and impact; RiskRover calculates risk scores and classes.

• Set testing intensity and strategy text, and select or remove project test techniques.

• Set effort-estimation baseline hours, size and complexity; RiskRover calculates effort.

• Create or update project roles and hourly rates for budget planning.

• Set activity and role allocations for each test level.

• Assign or remove quality attributes for individual PBS Components or clearly identified groups.

Review planning changes:

Use RiskRover MCP to review the PRA, Test Strategy, Effort Estimation and
Budget planning inputs for project 12345, including roles, hourly rates, and
activity and role allocations. Draft the requested changes, validate the
values, and show every proposed change. Do not submit anything until I
explicitly approve it.

For PBS Component quality assignments:

Use RiskRover MCP to review the current PBS quality assignments for project
12345. Add Security to every Component under Commerce > Checkout while
preserving its other qualities. Show every exact addition and removal and do
not submit anything until I explicitly approve it.

Removing an assigned quality also deletes its stored PRA, strategy, estimation and dependent-link data. Saving changes requires a connection with write access, explicit review and approval, and manual verification in RiskRover. For manual-token connections, use read-and-write scope.

For large projects, the agent reads planning data in manageable sections. Before saving approved changes, RiskRover rechecks the selected Components or rows.

Repository and workspace scope

RiskRover project access depends on your connection credentials, permissions and the requested project, not on the name or layout of the repository open in VS Code. A full application repository, a separate automation repository, and a test-only repository are all valid scopes.

The AI agent can assess only the workspace folders you deliberately opened. If relevant source or tests live elsewhere, use File > Add Folder to Workspace... to create a multi-root workspace, then start a new assessment. The agent should state which roots it inspected and identify missing roots as a limitation; it should not inspect unrelated directories implicitly.

The MCP server itself never reads repository files or runs commands. Workspace inspection is performed by the selected AI client using that client's permissions.

Authentication results

• Connection succeeded: RiskRover accepted the browser connection or manual token, and the configured project was checked when present.

• HTTP 401: the credentials are invalid, expired or revoked. For a browser connection, run Sign In again. For a manual token, run Set API Token with a current personal token.

• HTTP 403: check your connection’s access level and your membership of the requested project.

• HTTP 404: check the project ID and whether that user can see the project.

• Service unreachable or certificate error: check connectivity and corporate proxy or firewall settings, then repeat the appropriate connection check. Manual-token setup can optionally save without verification after explicit confirmation; this does not establish a successful connection.

A server shown as Running has completed MCP startup. Use RiskRover MCP: Test IDE Connection for a browser connection or RiskRover MCP: Test Connection for a manual token to check RiskRover authentication and project access.

Commands

Data and write permissions

• Read-only access supports automated test coverage assessment and drafting across PBS, PRA, Test Strategy, Effort Estimation and Budget planning.

• Write access is required to import a reviewed PBS, change Component quality assignments or save reviewed planning changes. Manual tokens need read-and-write scope.

• All changes require explicit review and approval. Replacing an existing PBS requires additional confirmation. Verify saved changes in RiskRover.

• Repository content is not sent to RiskRover. The selected AI provider may receive RiskRover data and workspace excerpts according to that client's data controls.

• The extension has no telemetry of its own and does not persist PRA or repository content.

Support: contact@riskrover.io