How minware's Security Model Differs from Cloud-Only Engineering Analytics Tools

All Posts
Share this post
Share this post

TL;DR: minware is cloud-based by default, like the cloud-only engineering analytics tools it competes with, and almost all customers use that standard connection. Its security model diverges in one option: an on-premise ingest agent that keeps source-system credentials and direct API access inside your own environment, then lets you inspect exactly what gets uploaded before it reaches minware’s cloud. Source code is hashed during ingest and never stored. minware is SOC 2 Type 2 compliant for the security criterion only, and the on-premise agent reduces access risk rather than the amount of data collected by default. The agent is optional and available on minware's Enterprise plan.

The security question in engineering analytics is whether you have to hand over credentials to your source systems before you can get any analytics at all. Cloud-only engineering analytics tools typically connect their cloud directly to your version control, project management, and CI/CD systems, which means granting access through OAuth or vendor-managed credentials. minware offers an optional alternative for security-sensitive teams: an on-premise ingest agent that runs inside your environment using your own credentials and uploads data to a bucket you can inspect. Source code is hashed during ingest and never stored, with or without the agent. This article walks through exactly what that model changes, what it does not change, and how to verify the claims for yourself.

How minware protects your proprietary code data

The minware security model centers on a single architectural decision: where your credentials live and what you can verify before data leaves your network.

What information minware ingests

minware ingests data from these categories of source systems to power its analytics reports:

  • Version control metadata: Commits, pull requests, branches, reviews, and merge activity from systems like GitHub, GitLab, Bitbucket, and Azure DevOps.

  • Project management fields: Tickets, statuses, estimates, epics, sprints, and the full history of every field where the source provides it, from tools like Jira, Linear, Azure Boards, and GitHub Issues.

  • CI/CD pipeline events: Deployment runs, build outcomes, and pipeline timing from systems like GitHub Actions, GitLab, and Azure Pipelines.

  • AI tool usage and token spend: Agent sessions, prompts, and cost data from Claude Code, Cursor, GitHub Copilot, and Codex.

  • Calendars: Calendar data from tools like Google Calendar.

  • Custom data sources: Data from other systems your team connects.

Source code content is hashed during ingest and never stored in minware's systems, though non-code files like agent instructions markdown or package dependency files may be stored to provide insights related to file content. What minware stores is metadata about development activity.

Security trade-offs in SaaS analytics

As a SaaS platform, minware requires some data to leave your environment. The question is what access you have to grant and what you can verify. Cloud-only tools require direct API access from their cloud to your source systems. minware's default is the standard cloud connection, which almost all customers use, so there is nothing to run locally. The optional on-premise agent adds setup effort but removes the need to grant minware direct credentials, and it exists for the security-sensitive minority that needs it. The agent requires setup on your end, and you are responsible for running it on a regular schedule. That setup cost is real, and it buys you a specific control: credentials stay behind your firewall.

Use cases for default cloud security

The standard cloud connection is sufficient for teams with no restriction on vendor API access, teams that want the fastest possible setup, and teams whose security review accepts SOC 2 Type 2 with read-only tokens. The on-premise agent is opt-in, and it is the exception rather than the rule: most minware customers run on the standard cloud connection. If your security model permits granting a vendor read-only OAuth access to your source systems, the standard connection works and requires no local configuration.

The hidden risks of cloud-only data pipelines

Cloud-only engineering analytics tools follow a consistent access pattern: the vendor's cloud connects directly to your source systems using credentials you provide.

Direct API access to source systems

Cloud-only tools require vendor-managed credentials or OAuth grants that let the vendor's cloud pull data directly from your version control, project management, and CI/CD systems. The risk is structural: if the vendor's cloud is compromised, the attacker has the same access the vendor had. In 2022, attackers used stolen OAuth tokens issued to Heroku and Travis CI to download data from dozens of organizations on GitHub.

Keeping access tokens behind your firewall

minware's on-premise model inverts that pattern. The on-premise ingest agent runs inside your environment and connects to source systems using your own credentials. API keys and direct source-system access never reach minware. The agent uploads processed data files, not live credentials. If minware's cloud were compromised, the attacker would find no source-system credentials to exploit because those credentials never left your network.

Securing data pipelines in managed environments

Managed environments like GitHub Enterprise Cloud and Jira Cloud still require some form of access. The on-premise agent connects to those systems from inside your network, so the connection originates from your environment rather than from minware's cloud. The agent authenticates to your managed systems using credentials you configure locally, and the same inspection controls apply to the uploaded data.

Securing minware data with local ingest agents

The optional on-premise ingest agent is the differentiator for security-sensitive teams in the minware security model, while the default cloud connection serves everyone else.

On-premise ingest for security control

The on-premise ingest agent is a local process that connects to source systems using your credentials, normalizes data locally, and uploads data files to a shared bucket. The primary benefit is access control: API keys and direct source-system access never reach minware. The data collected is the same as a standard connection by default. Any minimization is a consequence of the inspection control rather than the reason the agent exists. The agent is available on minware's Enterprise plan.

Local authentication for data ingest

The agent authenticates to your source systems using credentials you configure and store in your environment. minware never receives those credentials. The agent uses read-only tokens where the source system supports them, and all access tokens and personal access token (PAT) installations are read only. You vet and control updates to the agent, which means you decide when and how the software running inside your network changes.

Minimizing data exposure during ingestion

The agent handles these steps locally before any data leaves your environment:

  1. Processing: The agent retrieves data from your source system APIs and processes it locally.

  2. Anonymization: Sensitive fields can be omitted or anonymized using a hash, depending on how they are used.

  3. Preparation: The agent assembles data files for upload to the shared bucket.

The agent can limit minware's read access to exclude unused fields or the original values of anonymized fields, which lets you follow the principle of least privilege.

Audit trails for ingested files

The shared bucket is the verification mechanism that replaces trust with inspection. After the agent runs, you can review the raw data that minware will receive in a private staging bucket before sharing it with minware if you'd like. You can confirm that only the expected data is included and flag discrepancies before ingestion proceeds.

Source code hashing: what it protects and what it doesn't

Source code hashing is a specific privacy mechanism with specific boundaries.

minware hashing: data privacy mechanics

minware hashes source code during ingest so that code content is never stored in minware's cloud. The hash is a one-way transformation, so minware never retains the code itself.

What stays private after hashing

Code content stays private. minware does not store the original code. However, minware retains detailed metadata about development activity: commit timestamps, author information, pull request lifecycle stages, review activity, and related engineering metadata. This metadata is what powers cycle time, PR lead time, and delivery metrics. The distinction is between code content (protected) and code metadata (stored).

Balancing privacy and data utility

Hashing protects code content but means minware cannot show you code-level detail in reports. minware calculates cycle time, PR lead time, bug rate, and DORA metrics from metadata about when work happened and how it moved through your pipeline. This is a deliberate design choice: minware trades the ability to display code content for the guarantee that code content is never stored in minware's systems.

How minware secures your engineering tooling data

Access scoping is the second layer of the minware security model. Every integration uses read-only tokens, which limits what minware can do with your source systems even in the standard cloud connection.

Read-only access scoping for source systems and project data

Read-only access means minware cannot modify your source systems, under both the standard cloud connection and the on-premise agent model.

Project management integrations with Jira, Linear, Azure Boards, and GitHub Issues use read-only access to tickets, statuses, estimates, and custom fields. minware cannot create, edit, or delete tickets. minware ingests the full history of every field where the source provides it, which enables metrics like ticket cycle time broken down by status, but the access itself is read-only.

Managing CI/CD token access

CI/CD integrations read pipeline events and deployment data from systems like GitHub Actions, GitLab, and Azure Pipelines. Read-only access means minware cannot trigger pipeline runs, modify build configurations, or change deployment settings. The integration observes pipeline activity without the ability to affect it.

Verifying data integrity via inspectable buckets

The inspectable upload bucket is the mechanism that lets you verify what minware receives before ingestion.

How to audit your uploaded data

After the agent runs, you can inspect the shared bucket to see exactly what files were uploaded. The bucket contains the normalized data files the agent prepared, and you can open them to confirm the data matches what you expect. You can verify that no unexpected fields are included and that the volume of data aligns with the source systems you connected.

Audit data before it leaves your network

The inspection happens at the bucket, which is the boundary between your environment and minware's cloud. The agent can output data to any storage bucket, including a private staging bucket you control. This gives you a pre-ingestion audit point that a direct cloud connection does not offer.

Step-by-step data validation process

The validation process follows five steps:

  1. Run the agent: Execute the on-premise ingest agent with a private staging bucket as its output.

  2. Open the staging bucket: Navigate to the storage location where the agent uploaded files.

  3. Review the uploaded files: Inspect the data files to see what fields and records are included.

  4. Confirm the fields match your expectations: Verify that only the expected data categories are present.

  5. Share or hold the files: If the files match your expectations, move them to the bucket minware reads from. If you find unexpected data, hold them back and investigate before any data reaches minware.

Beyond minware: security responsibilities for you

The on-premise agent controls access, but it does not change what minware stores or where it stores it.

Security limits of the ingest agent

The on-premise agent does not reduce the amount of data collected by default. The data uploaded is the same as with a standard connection. The agent does not address data residency, GDPR, or Health Insurance Portability and Accountability Act (HIPAA) requirements. It runs only the ingest step inside your environment, so the uploaded data still resides in minware's cloud, and it does not change what minware stores. The agent's benefit is access control and verification.

SOC 2 Type 2 compliance scope

minware is SOC 2 Type 2 compliant for the security criterion only, audited for the most recent year ending in September. minware is not ISO 27001 certified. minware handles Protected Health Information (PHI) in compliance with HIPAA and can sign a Business Associate Agreement (BAA) for enterprise customers. Custom data residency is available on minware's Enterprise plan, while Professional runs on multi-tenant hosting with U.S. data residency.

When you need a fully on-premise tool

Some organizations need data to never leave their environment at all. minware is a SaaS platform. The on-premise agent keeps credentials local and lets you inspect uploads, but the data ultimately resides in minware's cloud. For organizations with a hard requirement that data never leaves their network, a fully on-premise analytics solution is required. minware is not the right tool for that case.

On-premise vs. cloud: security model trade-offs

The choice between minware's standard cloud connection, its optional on-premise agent, and a cloud-only vendor's direct API access is a trade-off between setup effort and control surface.

Restricting data exposure at the source

Cloud-only tools are simpler to set up but require you to grant their cloud direct access to your source systems. The on-premise agent adds setup effort but keeps credentials in your environment and gives you an inspection point before data leaves your network.

Controlling data exposure for analytics

The agent gives you three specific controls in practice:

  • Credential control: You choose which credentials the agent uses, and those credentials never leave your environment.

  • Inspection control: You can review exactly what the agent uploads before minware ingests it.

  • Revocation control: You can stop the agent and revoke the read-only tokens in your source systems, which cuts off minware's access.

Cloud-only tools offer revocation through OAuth token management. Because their cloud pulls data directly, there is no customer-side bucket to inspect before ingestion.

Self-hosted vs. cloud setup trade-offs

The table below compares the two models across five attributes:

Attribute minware on-premise agent Cloud-only vendor
Setup effort Higher (you run the agent on a regular schedule in your infrastructure and handle error management) Lower (OAuth connection, with credentials managed by the vendor)
Credential location Customer environment Typically vendor-managed (OAuth tokens or vendor-held credentials)
Inspection point Shared bucket (pre-ingestion review) Varies by vendor
Data residency U.S. by default, custom on minware's Enterprise plan Varies by vendor
Source code handling Hashed during ingest, never stored Not compared here (confirm in each vendor's documentation)

The minware on-premise vs competitors comparison comes down to where you want the control surface to live. Cloud-only vendors handle the infrastructure, which reduces your operational burden. minware's agent keeps the control surface inside your environment, which reduces your exposure if the vendor's cloud is compromised.

Choosing the right connection model

minware's on-premise ingest agent keeps source-system credentials inside your environment and gives you a bucket to inspect before data reaches minware's cloud. The data collected stays the same as a standard connection, so the agent fits teams whose security review rules out granting a vendor direct access. Teams comfortable with read-only OAuth access can use the standard connection with no local setup.

Start a 14-day free trial at minware.com, no credit card required, and connect your first data source.

FAQs

Does the on-premise agent reduce the data minware collects?

Not by default. The data collected is the same as a standard connection. The agent reduces access risk by keeping credentials in your environment, while the amount of data uploaded stays the same as a standard connection.

Can I use minware if my data cannot leave our environment at all?

No. minware is a SaaS platform and requires data to be uploaded to its cloud. The on-premise agent keeps credentials local but does not change where the data ultimately resides.

How long does it take to set up the on-premise ingest agent?

Setup requires configuration on your end, and the time it takes depends on your environment. minware's customer success team can guide the process.

Is minware ISO 27001 certified?

No. minware is SOC 2 Type 2 compliant for the security criterion only. It is not ISO 27001 certified.

What happens if I need to revoke minware's access immediately?

Stop the on-premise agent and revoke the read-only tokens in your source systems. Revoking the tokens cuts off minware's access, because the agent uses credentials that you hold and control.

Key terms glossary

On-premise ingest agent: A local process that runs inside your environment, connects to source systems using your own credentials, and uploads normalized data files to a shared bucket for minware to ingest.

Source code hashing: A one-way transformation applied to source code during ingest so code content is never stored in minware's cloud.

Read-only token: A credential that grants minware read access to your source systems without permission to create, modify, or delete data.

Inspectable upload bucket: A shared storage location where the on-premise agent uploads data files. You can review the files to confirm exactly what data is included before minware ingests it.

SOC 2 Type 2: An audit report covering minware's security controls over a period of time. minware's report covers the security criterion only.