Custom internal tools and portals,at a fixed price.

Get admin panels, customer portals and back-office apps built with tests and screenshots on every pull request.

Pento builds the order portals, approval flows and dashboards that usually sit behind the product roadmap or live in spreadsheets and SaaS seats, on Steer by Pento, the platform we run every project on. Forward deployed engineers write the spec and agents build on branches in your GitHub.
Illustrative Steer project building a customer order portal. Five goals (sign in with SSO, order history, reorder in one click, invoice download, and an admin screen to approve accounts) move from Ready to In progress to Merged to Verified, two at a time. As each goal merges, its screen appears in the portal preview and a screenshot is attached to its pull request.

A fixed price, or a team

A tool with a clear scope gets one price, in writing before any code exists. Several tools, or one that keeps growing, can go to a Pento team of forward deployed engineers.

Code in your GitHub

It lives there from the first commit, in a stack your engineers can run, with no per-seat license to renew.

Wired to your systems

Each tool is built against what you already run (your identity provider, your database, your ERP or CRM) through their own APIs.

What is custom internal tool development?

Custom internal tool development is building the software a company runs on behind the scenes: admin panels, back-office apps, approval flows and dashboards, plus the customer, partner and order portals that sit on top of them, shaped to your process instead of a vendor's template. Pento builds them on Steer by Pento, as fixed-price projects when the scope is clear, and the code lives in your repositories.
Who does what on an internal tool project
  1. 01Interview
    • YouYour ops lead or account manager takes the interview
    • Steer + AI agentsWrite the interview, keep the answers as context
  2. 02Spec
    • YouAgree the spec and the price before any code
    • Forward deployed engineersWrite the spec: one goal per screen or flow
    • Steer + AI agentsOrder the goals by what depends on what
  3. 03Build
    • Steer + AI agentsBuild screens in parallel, each in a fresh sandbox
    • Review + checksTests and a screenshot on every pull request
  4. 04Review
    • YouReview the screen itself, not a description
    • Forward deployed engineersJudge each change
    • Review + checksA model that did not write the code checks it
  5. 05Merge
    • YouApprove changes that touch data or external systems
    • Review + checksLow-risk changes merge on a green review and CI
  6. 06Handoff
    • YouYour engineers inherit the code and its history
    • Forward deployed engineersCan keep building it with your team, monthly

Changes requested in review go back to Build.

Why they never get built

The tools that run the business are always next quarter.

Product engineering owns the roadmap, so the order portal, the approval flow and the admin screen your ops team asks for keep slipping. Meanwhile the work happens in spreadsheets, shared inboxes and a growing stack of SaaS seats. Low-code tools are a fair answer for simple forms. They get harder to live with once a tool needs your SSO, your data model and real tests.
Back office

The admin tool your ops team keeps asking for

TodayRoadmap backlog · spreadsheet workaround · next quarter
On SteerFixed price upfront · code in your GitHub

Refunds, account changes, catalog fixes and approvals done by hand in a database console or a spreadsheet. We build the admin panel around how your team already works, with roles, an audit log and tests, without pulling your product engineers off the roadmap.

See the evidence on every PR →
Customer & order portals

The B2B portal your customers expect

TodayOrders by email · PDF price lists · a per-seat portal add-on
On SteerSelf-serve ordering · your SSO · connected to your ERP

B2B buyers want to see order history, reorder in one click and download invoices without calling anyone. We build the portal on the systems that already hold the orders, so customers see the same numbers your team does.

See ERP integrations →
Low-code and SaaS seats

When the low-code app hits its ceiling

TodayPer-seat licenses · the vendor's data model · logic you can't test
On SteerPlain code · your data model · tests on every change

Low-code platforms are quick for a form over one table, and we will say so when that is all you need. Once the tool needs your SSO, several systems, custom permissions and real tests, a small codebase your engineers can read is easier to live with than another seat tier.

See how pricing works →

✳ Scheduling, quoting, inventory views and partner portals follow the same pattern. A first conversation shows where to start.

How it runs on Steer

One goal, one session, one evidence report.

Steer by Pento splits the tool into goals, each with acceptance criteria and a checklist. Each goal runs in its own session: an agent in a fresh, isolated sandbox reads the spec, writes the code, runs the tests and attaches screenshots, then an independent review model and a forward deployed engineer judge the change. Independent goals run in parallel.
  1. 01

    Interview the people who use it

    When a goal needs a conversation, Steer writes the interview and your ops lead or account manager takes it in the browser. The answers become versioned context the spec builds on.

  2. 02

    One goal per screen or flow

    Sign-in, order history, reorder, invoices, approvals: each is a goal with acceptance criteria, ordered by what depends on what, so the independent ones run at the same time.

  3. 03

    A screenshot on every pull request

    Anything on screen comes back as a screenshot attached to its pull request, next to the tests that ran, so your team reviews the screen itself, not a description of it.

  4. 04

    Verified on your default branch

    A merged goal counts as verified only when CI is green on your default branch.

Illustrative Steer session for one goal of a customer order portal: the admin screen that approves accounts. An agent reads the spec, edits the screen, runs the suite, attaches a screenshot to pull request 221, and a review model confirms the change conforms to the spec. The change touches account data, so the run stops for a person to approve it.
Evidence

Every pull request arrives with its evidence.

Every pull request carries an evidence report: what changed, which acceptance criteria it meets, the tests that ran, screenshots of anything on screen, and the verdict of a review model that did not write the code.
  • Merges follow your policy

    Low-risk changes merge on a green review and CI. Changes that touch data or external systems ask a person first, at the autonomy level you set.

  • A handoff your engineers can run

    The code, its history and the evidence stay in your GitHub, so your engineers inherit the tool with its reasoning written down.

Illustrative evidence report for one pull request, PR #217, Reorder in one click: three acceptance criteria confirmed by the review, a screenshot of the order list with its new Reorder buttons, 36 tests passed, an approve-with-findings verdict with one low-severity UX finding from a review model that did not write the code, CI passed, and a low-risk merge under the project's policy.
What you get

What every internal tool project includes

The tool, and everything your team needs to trust it and keep it. The scope is written down before the price is, and the price before any code.

Every internal tool or portal project ships with:

A written spec with acceptance criteria for every goal, agreed before any code exists
A written quote before the build starts: a fixed price for a clear scope, or a monthly fee for a Pento team
Code in your GitHub from the first commit, in a stack your engineers can run
Sign-in through your identity provider, with roles and an audit log where the tool needs them
An evidence report on every pull request: tests, screenshots and the review verdict
A seat for your team in the Steer workspace from day one
Stories

Internal tools shipped on Steer.

Client names stay private. The shape of the work doesn't.
Case studies →

3 weeks

A 3-week internal admin tool, ~120 PRs

Agents executing in parallel, engineers reviewing evidence on every pull request.

8 days

An 8-day MCP integration into a client's CRM

A custom agent workflow wired into systems that were never designed for agents, with end-to-end behavioral tests included.

How we engage

Two ways to work with us. Neither is hourly.

Both start with the free audit, which sets the scope and the shape of the team.
Monthly · set by pod size

A team we build for you

Pento forms a pod of forward deployed engineers for your work, for example four engineers or whatever machine learning, software, DevOps and MLOps talent you need. They work with Steer alongside, and the fee depends on the pod's size.

Best for
An ongoing roadmap, a large project such as a migration, or scope that will change as you learn
Who carries the risk
Shared: you set the priorities, and we own how the work is built and the evidence that it's done
Talk to us about a team
One price · a clear scope

A fixed-price project

For integrations and smaller projects you can scope up front. After the free audit we write the scope with you as goals with acceptance criteria and fix the price. Evidence on every pull request shows when each goal is done.

Best for
Integrations and smaller projects with a clear scope. First projects ship in weeks
Who carries the risk
Pento: the price is fixed, so an overrun is ours to absorb
Get a fixed-price quote
How pricing works →
FAQ

Questions about custom internal tools

Not covered here? Bring the question to a discovery call. An engineer reads every request.

Book a discovery call

Contact us

Let's build your internal tool on Steer

Tell us which admin panel, portal or approval flow is stuck behind the roadmap. We start with a free audit, then quote a fixed price for a clear scope, or a Pento team of forward deployed engineers for a longer run of work. An engineer reads every request.