Introducing the Leera MCP server: connect your AI client

A useful AI workflow needs the project context behind the task. Leera’s built-in Model Context Protocol server gives compatible AI clients access to supported workspace operations, so an assistant can retrieve current records and perform authorized actions without relying only on a pasted summary.

Bring project context into the tools you use

A coding assistant can read an issue’s acceptance criteria before implementation, update the issue as work progresses, and write the implementation notes in a document. With the required scopes, it can create QA cases, work with plans and runs, and record results and evidence from a tester or automation system.

These are example workflows, and they require the relevant client support, tool access, and configuration. Native automation tools can validate and save case scripts and queue automated run items, but execution requires compatible runners and an environment configured for your deployment. Creating a case or saving a result does not execute a test. Inspect the run and evidence before relying on the outcome.

Choose access for the assignment

Leera exposes MCP over Streamable HTTP with OAuth 2.1 or scoped personal access tokens. The method you use depends on the client and deployment. New personal tokens default to Planner read access. OAuth consent can request broader scopes, including writes, so review the selection before connecting. Restrict projects and grant only the capabilities the task needs. Set an appropriate expiry and revoke credentials that no longer have a purpose.

A token’s plaintext secret is shown when it is created. Store it in the client’s supported credential mechanism and give the connection a meaningful name. A read-only sprint summary should not need the same access as an integration responsible for creating project records.

Actions use the identity and permissions of the person who authorized the connection or created the token. Once access is granted, a permitted write can change a record directly; Leera does not add an approval gate before every action. Check the resulting issue history, document changes, and QA evidence. The connected client may also offer its own confirmation controls.

Check the available tool catalog

The supported surface includes project information and operations around issues, sprints, documents, and QA. What a specific connection sees depends on the exposed tools, enabled modules, and credential access. Do not assume every web control has an equivalent tool or that every client presents the same catalog.

Read the tool’s requirements and verify important results in Leera. For an action that creates a record, ask the client to return its identifier and read the saved record back. That makes the outcome easier to inspect.

Start with your actual endpoint

For parallel coding sessions, the agent manager setup guide explains documented MCP configuration paths in Conductor, Superset, RunPane, and Nimbalyst. Configure access in the environment where the selected agent runs, and verify a fresh worktree or session. These combinations have not yet been validated end to end with Leera.

Copy the endpoint from your Leera settings and follow your client’s current connection instructions. A sample configuration for one client is not a universal format. For a self-hosted instance, make sure the public reverse proxy forwards /mcp to the backend alongside /api.

Test a familiar issue first, then perform one deliberate change in an appropriate trial project. Check the workspace, project, and saved result before expanding the workflow. The complete MCP guide includes example prompts, an access checklist, and troubleshooting steps.