A Model Context Protocol Server for GitLab: Exposing Flaky Job Signals and Merge-Request Metrics
Bibliographic record
Abstract
The Model Context Protocol (MCP) is a resource-oriented standard that enables client applications to request structured contextual data from external systems. In simple terms, MCP functions as a consistent interface where client tools can “order” specific information items, and the server returns them in a structured format. This research project developed an MCP server that integrates with GitLab APIs to expose software engineering signals as consumable MCP resources, facilitating downstream analysis by model-driven clients such as Cline, Roo, or Claude Code. Two primary categories of MCP resources were implemented. The first identifies flaky jobs within continuous integration pipelines by detecting instances where jobs fail and later succeed on the same code revision, providing structured metadata including commit identifiers, pipeline context, and job logs. The second category extracts merge-request heuristics and metrics, such as time-to-first-review, time-to-merge, comment activity, reviewer overlap, and code change size, enabling consistent access to project-level development signals. Several limitations of MCP were observed. Straightforward improvements include extending GitLab endpoint coverage, refining caching strategies, introducing webhook-driven updates to reduce data staleness and hardening authentication for enterprise environments. Long-term challenges involve real-time event streaming at scale and exposing GitLab POST endpoints in a safe manner, possibly for use cases like code review (look at merge requests and approve or make suggestions) or even authoring new code. The findings indicate that MCP can effectively serve as a standardized interface for exposing targeted software engineering resources, while highlighting future directions in scalability, accuracy, and enterprise readiness.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.004 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.002 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.003 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".