MCP server for Codebeamer ALM with read and write capabilities
The Codebeamer MCP server demonstrates solid fundamentals with clear naming conventions, comprehensive parameter schemas, and good pagination support. All 20 tools follow verb_noun naming (list_, get_, create_, update_, add_, search_). Input schemas are consistently present with typed parameters and descriptions. However, several critical gaps limit production readiness: output schemas are not documented in the source, error handling lacks recovery guidance, tool annotations (readOnlyHint/destructiveHint) are absent, and descriptions sometimes lack sufficient context about prerequisites or when to use each tool over similar alternatives. The server covers a good breadth of Codebeamer operations (projects, trackers, items, comments, associations, risks) but lacks patterns for defensive design (e.g., no dry-run for destructive operations, no user confirmation step for writes).
Add a comment to a Codebeamer work item. Supports plain text and wiki markup formats. Returns the created comment.
Create an association (link) between two Codebeamer work items. Common association types: 'depends on', 'blocks', 'related to', 'derived from'. Use get_item_relations on an existing item to discover valid association type IDs.
Create a new work item in a Codebeamer tracker. Supports setting name, description, status, priority, assignees, story points, and custom fields. Returns the created item.
Add a downstream reference from one Codebeamer item to another. Downstream references represent derivation/traceability links (e.g. a requirement derived from another). The 'from' item gets the downstream reference pointing to the 'to' item.
Create a risk item in a Codebeamer risk tracker. Requires name and a risk tracker ID. Supports setting description, probability, impact, and mitigation plan.
Output schemas not documented. The source code shows input schemas but does not document what each tool returns. LLMs cannot plan multi-step workflows or extract required fields for downstream tools without seeing return structures.
Missing tool annotations. No readOnlyHint or destructiveHint declared. Destructive tools (create_item, update_item, add_comment, create_association, create_reference, create_risk) should explicitly mark themselves as having side effects. Agents cannot distinguish safe-to-retry READ operations from WRITE operations that risk duplication.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 52 | 2026-07-28+ | v2 |
| 2026-03-09 | D | 58 | - | v1 |
Get full details of a Codebeamer work item including description, status, priority, assignees, custom fields and test steps.
Get all comments (discussion thread) for a Codebeamer item.
Get the full structured detail of a Codebeamer work item: project, priority, assignees, created/updated timestamps, story points, custom fields and test steps. Description is intentionally omitted — fetch it via get_item.
Get upstream and downstream traceability references for a Codebeamer item. Upstream references point to items this one is derived from (e.g. requirements). Downstream references point to items derived from this one (e.g. test cases).
Get all relations (associations) for a Codebeamer item. Shows incoming and outgoing links like 'depends on', 'blocks', 'derived from', etc.
Get all Review Hub reviews for a Codebeamer tracker item. Shows the overall review result (APPROVED/REJECTED/UNDECIDED), individual reviewer votes, and review configuration (required approvals/rejections).
Get detailed information about a specific Codebeamer project.
Get details about a specific tracker including its type, project, description and fields.
Get the field schema for a Codebeamer tracker. Shows all available fields, their types (text, choice, user, etc.), and which are required.
Get details about a Codebeamer user including name, email, and account status.
List all Codebeamer projects. Returns project names, IDs, keys, and status.
List all items (work items) in a specific Codebeamer tracker. Shows ID, summary, status, priority and assignees. Returns pagination support.
List all trackers (bug trackers, requirements, test cases, tasks, etc.) in a Codebeamer project.
Search for Codebeamer items using Codebeamer Query Language (cbQL). Example: 'tracker.id IN (101) AND status.id = 10'.
Update an existing Codebeamer work item. Supports updating name, description, status, priority, assignees, story points, and custom fields. Only provided fields are updated.
No dry-run or confirmation step for write operations. Irreversible operations (create_item, update_item, add_comment, create_association, create_reference, create_risk) lack a safety mechanism. An agent mistake could create duplicate items or spam comments. Consider adding an optional 'dry_run' parameter or a separate confirmation step pattern.
Error handling lacks recovery guidance. No evidence in descriptions of actionable error messages that tell the LLM what to do next. E.g., 'User not found. Try search_users() first' or 'Invalid tracker ID. Call list_trackers() to discover valid IDs.'
Tool descriptions lack prerequisite context. For example, create_association description does not explain that the agent must first call get_item_relations to discover valid association type IDs. Similarly, create_risk and create_item do not hint that tracker/field schema discovery tools (get_tracker_fields) should be called first to find valid field/status/priority IDs.
get_item_details and get_item overlap in purpose. Descriptions do not clearly explain why an agent should call get_item_details over get_item, or vice versa. The distinction (get_item_details omits description but includes project/priority/timestamps/story_points/custom_fields) is documented but could be clearer for LLM decision-making.
search_items expects Codebeamer Query Language (cbQL) but does not document query syntax or provide examples beyond one parenthetical. LLMs will struggle to construct valid queries without detailed syntax documentation or a separate query helper tool.
No parameter bounds or enum constraints documented for numeric IDs and choice fields. For example, create_item accepts 'status', 'priority', 'categories' as numeric IDs but does not explain where to find valid IDs or document valid ranges. Agents must infer this from context.
customFields parameter in create_item and update_item lacks detailed documentation of expected structure. Description says 'Array of {fieldId, type, value} objects' but does not explain valid type values or how to validate/format the value for each type (text, choice, user, etc.). Agents will guess.