Middleware

Engineering analytics, honestly reviewed. No marketing fluff, just what you get when you self-host it.

TL;DR

• What it is: Open-source (Apache-2.0) DORA metrics platform — Deployment Frequency, Lead Time for Changes, Mean Time to Restore, Change Failure Rate, tracked automatically from your CI/CD and project management tools [1].
• Who it’s for: Engineering leaders — CTOs, VPs of Engineering, team leads — who are being asked to report delivery performance and refuse to pay enterprise SaaS prices to do it [1][2].
• Cost savings: LinearB, Jellyfish, and Swarmia all run $15–40/seat/month with annual contracts required. Middleware self-hosted runs on a single Docker container with zero per-seat cost [1][2].
• Key strength: Arguably the lowest-friction setup in the self-hosted engineering analytics category. One Docker command. Postgres and Redis bundled. Running in under 30 minutes [1].
• Key weakness: 1,571 GitHub stars and essentially no independent third-party reviews — this is an early-stage project you’d be adopting before the community has fully validated it.

---

What is Middleware

Middleware (middlewarehq.com) is an open-source engineering management platform built around DORA metrics — the four measurements that Google’s DevOps Research and Assessment program identified as the clearest signal of software delivery performance [1].

The four metrics are:

• Deployment Frequency: How often your team ships to production
• Lead Time for Changes: Time from commit to production deployment
• Mean Time to Restore: Time from incident detection to service recovery
• Change Failure Rate: Percentage of deployments that cause failures needing remediation

These four numbers have become the de facto KPI framework for engineering leaders since the Accelerate book and the subsequent DORA state-of-DevOps reports mainstreamed them. The problem is collecting them without a dedicated tool. Your options before Middleware were: build internal dashboards (expensive engineering project, then maintained forever), or pay enterprise SaaS rates. Middleware self-hosted removes the recurring fee from the equation [1][2].

The GitHub description is admirably plain: “Open-source engineering management that unlocks developer potential” [1]. The company’s website pitches it as helping teams “Deliver Better Software Faster, Without the Drama” [2] — which is either the most honest tagline in engineering analytics or the most understated, depending on how many all-hands slides you’ve had to present about CI/CD lag without data.

The Apache-2.0 license is meaningful. Unlike some tools that present themselves as open-source but restrict commercial use the moment you actually read the license, Apache-2.0 gives you unrestricted use, modification, and distribution rights — including in commercial contexts [1].

At 1,571 GitHub stars, Middleware is a legitimate project in early growth, not a weekend experiment. But it hasn’t reached the community depth of more established self-hosted tools, and that affects how much risk you’re taking on.

---

Why people choose it

Without substantial independent third-party reviews to synthesize — and notably, none were found during research for this article — the strongest case for Middleware is structural: it solves a specific, well-defined pain point for a specific buyer at a price that alternatives can’t match.

The engineering analytics SaaS market is crowded with expensive tools that serve the same organizational need. Middleware’s own website publishes comparison pages against LinearB, Jellyfish, and Swarmia [2], which are the three SaaS tools most often purchased in this space. None of them have meaningful free tiers. All of them quote on request, which in B2B SaaS translates to “not cheap.” For a 40-engineer team, these tools typically cost $600–$1,600/month — figures that fund roughly one engineer-month per year.

The website claims 2.6x faster bottleneck resolution, 87% reduction in time spent creating reports, and 3x better project predictability [2]. These are vendor-reported metrics, not audited outcomes, so treat them as signals about what the product optimizes for rather than benchmarks to hold them to. What they do tell you: Middleware is explicitly positioning against the manual reporting problem — the hours of engineering manager time currently spent pulling data into Google Sheets and making it look good for the board deck.

The honest version of the buyer profile: a 30–80 person engineering org where the CTO needs a DORA dashboard for leadership, the LinearB quote just came in at five figures annually, and there’s one DevOps or platform engineer who can manage a Docker container. That’s the decision tree that ends at Middleware.

---

Features

Based on the GitHub README and features page [1][3]:

Core DORA engine:

• Deployment Frequency tracked from CI/CD event data [1]
• Lead Time for Changes from commit to production [1]
• Mean Time to Restore from incident data [1]
• Change Failure Rate derived from deployment events [1]
• Automated collection — no manual data entry or spreadsheet imports [1]

Pull request analytics:

• PR cycle time, review time, and merge patterns [3]
• Code review throughput and bottleneck identification [3]

Planning and delivery:

• Sprint reports [3]
• Flow metrics from planning to delivery [3]
• Project predictability tracking [2]

Dashboards and reporting:

• Customizable dashboards [1]
• KPI visualization [1]
• Report automation [2]

AI features:

• “AI Insights” is listed prominently on the features page [3]
• What specifically the AI does, how it derives insights, and what models or techniques power it are not documented in publicly available materials reviewed — this could be meaningful or a marketing checkbox, and it’s worth asking the vendor directly before making it part of your rationale

Integration surface:

• CI/CD tool integrations [1]
• Project management platform integrations [1]
• REST API [1]

Notable gap: the README says “integration with various CI/CD tools” and “popular project management platforms” but doesn’t enumerate which ones [1]. Before committing to Middleware, verify that your specific stack — your CI system, your issue tracker, your incident management tool — is actually supported. Running the evaluation on your real data is the only reliable confirmation.

---

Pricing: SaaS vs self-hosted math

Self-hosted (open-source):

• Software license: $0 (Apache-2.0) [1]
• Infrastructure: $10–20/month VPS for a typical team
• Maintenance: your time

What the SaaS alternatives charge: LinearB, Jellyfish, and Swarmia don’t publish per-seat pricing — all three require a sales conversation. The general range for engineering analytics tools at this tier is $15–40/seat/month with annual contracts. Specific pricing data was not available for this review and was not fabricated.

Rough math for a 40-person engineering team at a conservative $20/seat/month: $800/month, or $9,600/year. Self-hosting Middleware on a $20 VPS: $240/year. The delta — around $9,360/year — funds roughly six weeks of engineering time. That’s the framing your CTO needs when justifying the evaluation time.

If Middleware offers a managed cloud option, that pricing was not available in materials reviewed for this article.

---

Deployment reality check

The setup story is Middleware’s strongest practical argument. The quick start from the README is a single Docker run [1]:

docker volume create middleware_postgres_data
docker volume create middleware_keys
docker run --name middleware \
           -p 3333:3333 \
           -p 9696:9696 \
           -p 9697:9697 \
           -v middleware_postgres_data:/var/lib/postgresql/data \
           -v middleware_keys:/app/keys \
           -d middlewareeng/middleware:latest

PostgreSQL and Redis are bundled in the container. The app comes up at http://localhost:3333. No separate database provisioning, no Redis configuration, no multi-container YAML to debug [1]. For self-hosted engineering tooling, this is genuinely minimal friction.

What you need for production:

• Docker installed on a Linux VPS (2–4GB RAM recommended)
• A reverse proxy (Caddy or nginx) and a domain for HTTPS access
• Firewall rules to restrict port exposure
• An SMTP provider if email notifications are needed

What can go sideways:

The single-container default bundles Postgres inside the container. This is fine for evaluation and internal use, but it means your database lives and dies with your container. For production, you should externalize Postgres with proper backups — the container approach gives you no isolation between the application failing and losing your historical metrics data [1].

Updates require stopping the container, removing it, and re-pulling, which means planned downtime [1]. There’s no zero-downtime upgrade path documented in the README.

No Helm charts or Kubernetes deployment guide appears in public documentation — teams that run everything on k8s don’t have a documented production path and would need to build one themselves [1].

Realistic time estimates:

• Local evaluation instance: 15–30 minutes
• Production with HTTPS and external Postgres: 2–4 hours
• For someone new to Docker: budget a full afternoon plus debugging time

---

Pros and cons

Pros

• Apache-2.0 license — genuinely permissive. Self-host, modify, embed in internal tooling, no commercial licensing conversation required [1]. LinearB, Jellyfish, and Swarmia are all closed-source SaaS.
• Fastest time-to-running in the category. One Docker command, bundled dependencies, app available at localhost:3333. For a self-hosted tool, this is an unusually low friction bar [1].
• Covers all four canonical DORA metrics. Not a partial implementation — tracks the complete framework that engineering leaders are expected to report [1].
• REST API included. Programmatic access to metrics means you can pull data into existing BI tools or executive dashboards [1].
• PR analytics built in. Pull request cycle time, review patterns, and bottleneck identification are core features, not an enterprise add-on [3].
• Zero recurring cost. For a team with a technical person to manage one container, the cost difference versus SaaS alternatives is measured in thousands of dollars annually.

Cons

• 1,571 GitHub stars is real but early. Comparable tools in adjacent self-hosted spaces (Plane, Gitpod, various observability tools) are in the 10,000–50,000 star range. Middleware is pre-breakout, and long-term maintenance continuity is an open question worth watching.
• No independent reviews found. LinearB has extensive G2 and Capterra coverage. Jellyfish has case studies. Middleware has essentially no third-party validation at this writing. This doesn’t mean the tool is bad — it means you’re absorbing that validation risk yourself.
• Integration specifics are underdocumented. “CI/CD tools” and “project management platforms” without a named list makes it hard to assess fit before deploying [1].
• Single-container Postgres is a production concern. Bundled database is great for setup, risky for data durability without explicit externalization and backup configuration [1].
• No Kubernetes docs. Teams running k8s production environments don’t have a documented path [1].
• AI Insights details are opaque. Prominently marketed but not specifically documented — hard to evaluate the actual value [3].
• No SSO or audit logging documented. For teams in regulated industries or companies with security review requirements, these are likely blockers.

---

Who should use this / who shouldn’t

Use Middleware if:

• You’re an engineering leader at a software company being asked to track and report DORA metrics and the quoted price for LinearB or Jellyfish made you open a new browser tab to search for alternatives.
• You have one technical person who can manage a Docker container and set up a reverse proxy. The install is genuinely simple; the maintenance responsibility is real.
• Your stack includes common CI/CD and project management tools — verify your specific integrations before committing, but the probability is reasonable.
• You’re comfortable adopting a product at its growth stage, with the understanding that you’re part of the early adopter cohort rather than buying a proven enterprise tool.
• Apache-2.0 licensing matters for compliance, internal tooling reuse, or resale reasons.

Skip it if:

• Your team runs zero infrastructure and doesn’t want to own a container. The install is easy; the ownership isn’t zero.
• You need guaranteed integration with a specific tool in your stack and can’t verify support through testing.
• You need enterprise auth: SAML SSO, LDAP, fine-grained RBAC with audit logs. These aren’t documented for the open-source version.
• Your security or compliance team needs a vendor with SOC 2 attestation, a DPA, and a legal entity to hold accountable. Middleware doesn’t offer that yet.
• You need a tool with five years of customer evidence and community troubleshooting resources. That’s not this product in its current form.

Consider a SaaS alternative if:

• Your engineering org is over 200 people, the metrics need to connect to business outcome reporting, and you need a vendor relationship with SLAs. LinearB or Jellyfish are the right call.
• You want something in between: Swarmia or Sleuth offer lighter-weight SaaS options with better pricing than the enterprise tier, and Sleuth has a free tier worth evaluating before choosing either direction.

---

Alternatives worth considering

From Middleware’s own comparison pages and the broader market [2]:

• LinearB — the best-resourced player in the space. Deep integrations, AI features, strong case study library. Closed-source SaaS, enterprise pricing. Use this if budget isn’t a constraint and you want a vendor with support infrastructure.
• Jellyfish — similar positioning to LinearB, stronger emphasis on connecting engineering investment to business outcomes. Better fit for larger organizations with complex cross-functional reporting needs.
• Swarmia — lighter and more affordable than LinearB/Jellyfish, still SaaS. Good entry point for teams that want tooling without the enterprise commitment. Not free, but not enterprise-priced either.
• Sleuth — SaaS DORA metrics tool with a free tier. Narrower scope than Middleware (four metrics, less broader analytics), but the free tier makes it worth evaluating before choosing self-hosted.
• Grafana + custom dashboards — many engineering teams build their own DORA dashboards in Grafana from CI/CD event data and GitHub API. Free, fully customizable, requires significant engineering time to build and maintain. Better for teams that already run Grafana for infrastructure monitoring.
• Haystack — engineering analytics SaaS focused on code review metrics and DORA. Closed-source, per-seat pricing.
• Pluralsight Flow (formerly GitPrime) — the most established name in this space, now owned by Pluralsight. Enterprise pricing, enterprise features, enterprise bureaucracy. Only relevant at 500+ engineers.

For the specific audience this review is written for — non-technical founders or first-time CTOs navigating this for the first time — the realistic decision is Middleware self-hosted vs. Swarmia or Sleuth cloud. If you have a technical person, Middleware eliminates the recurring bill. If you don’t, Swarmia or Sleuth with a free trial is a lower-risk starting point.

---

Bottom line

Middleware solves a real and common problem with an unusually low setup bar. Engineering leaders need DORA metrics; the tooling market charges enterprise prices to provide them. Middleware self-hosted — Apache-2.0 licensed, one Docker command, bundled dependencies — eliminates the recurring cost entirely. The tradeoffs are honest: 1,571 GitHub stars puts this in early-growth territory, the integration documentation leaves gaps you’ll need to verify, and bundling Postgres in the default container is a data durability risk you’ll need to manage consciously in production. None of those are disqualifying, but they’re the things to evaluate before treating this as critical infrastructure. If your current alternative is paying $8,000–$15,000/year for LinearB or spending two months building internal dashboards, Middleware is worth an afternoon to test against your real data.

If the Docker setup is the blocker, that’s exactly what unsubbed.co’s parent studio upready.dev deploys for clients. One-time fee, done, you own the infrastructure.

---

Sources

1. GitHub — middlewarehq/middleware (1,571 stars, Apache-2.0 license). Open-source engineering management that unlocks developer potential. https://github.com/middlewarehq/middleware
2. Middleware — Official Website (middlewarehq.com). Engineering productivity software for delivery leaders. https://middlewarehq.com/
3. Middleware Features Page — DORA metrics, PR insights, flow metrics, AI Insights. https://middlewarehq.com/features