Turn failing CI/CD pipelines into 30-second explanations.
Exlogare reads your pipeline logs, finds the root cause, and posts a fix-ready analysis straight to your merge request. Connect GitLab in under 5 minutes.
No credit card · GitLab self-hosted & gitlab.com · OpenAI or Qwen
$ pytest -q
FAILED tests/test_billing.py::test_charge_retry
E sqlalchemy.exc.OperationalError: (psycopg2.OperationalError)
E connection to server at "db" (10.0.0.7), port 5432 failed
E FATAL: password authentication failed for user "exlogare"
...47 more lines truncated...
Postgres auth rejected — `DB_PASSWORD` changed, secret not rotated in CI.
The job test connects as exlogare using $DB_PASSWORD. The CI variable was last updated 3 weeks ago; the database password was rotated yesterday (commit a74f3e0).
Pipeline failures steal more time than the bugs themselves.
Every team loses hours every week scrolling through logs to explain the same categories of failures over and over.
Hours lost to "why did this fail again?"
Your senior engineers are the default on-call for every red pipeline. That's the wrong bottleneck.
50,000 lines of logs, one real culprit
Stack traces, retries, flaky tests, broken images — the signal is buried. Searching is slow, guessing is expensive.
Midnight pings on the same class of error
Nobody writes a postmortem for the 11th "cache layer missed" — it just wakes someone up again.
Three steps. Under five minutes.
Connect once. From then on, every red pipeline gets a human-readable RCA inside the merge request it broke.
- 01
Connect GitLab
Webhook, OAuth app, or personal access token — whichever matches your governance. Works with gitlab.com and every self-hosted edition.
- 02
A pipeline fails
Exlogare fetches the failed job logs, strips noise, and hands the relevant window to the LLM of your choice (OpenAI or Alibaba Qwen).
- 03
The fix lands in the MR
A structured analysis — likely root cause, offending file/line, suggested patch, and confidence — posted as a merge-request comment in under 60 seconds.
Built for the real flow of a DevOps team.
Nothing decorative. Every feature below exists because a DevOps team asked for it on day one.
AI root-cause analysis
Swap between OpenAI and Alibaba Qwen per tenant. Logs are chunked, deduplicated, and scored — you only pay for what matters.
GitLab webhook + OAuth
Self-hosted GitLab? OAuth with polling. Locked-down network? Webhook only. Want both? Hybrid mode, same pipeline.
Jenkins support
Drop a small post-build step and Exlogare ingests the build log the same way it ingests GitLab jobs.
Slack · Telegram · Matrix
Route RCAs to the channel or group that actually owns the failing service — not to a generic #ci-alerts firehose.
Pipeline trends
See which failure classes are getting more frequent, which jobs are flakiest, and which fixes actually stuck.
Self-hosted ready
Docker Compose, one VPS or four, your Postgres and Redis. Your logs never leave your own pipe unless you want them to.
Answers to the questions every DevOps lead asks first.
Can we self-host Exlogare?
Yes. The entire stack (FastAPI + Postgres + Redis + Celery + React SPA + Nginx) ships as Docker images and Compose files. Bring your own LLM API key.
Which LLMs are supported?
Today: OpenAI (GPT-4o/mini family) and Alibaba Qwen via the DashScope OpenAI-compatible endpoint. The LLM layer is pluggable, so adding a new provider is a small adapter.
Where does our log data live?
On the infrastructure you connect. Exlogare only stores the redacted excerpts it sends to the LLM plus the RCA text. You can self-host to keep 100% of the data inside your VPC.
Does it work with Jenkins?
Yes. A small post-build script posts the build log to Exlogare; the rest of the pipeline is identical to the GitLab flow.
Is there a free tier?
Yes — 50 RCA analyses per month on the Free plan, no credit card. Usage-based pricing kicks in after that.
How do you handle secrets in logs?
Log chunks are passed through a redaction layer (tokens, JWTs, AWS keys, GitLab tokens, basic-auth URLs) before anything is sent to the LLM. The raw log stays on disk inside your infra.
Set up in 5 minutes. Free for the first 50 analyses.
Connect your first GitLab project, break a pipeline on purpose, watch the RCA land in the MR. That's the entire demo.