How the Mirego team uses Forra
Eight applications built by developers, project managers, a QA specialist and a marketer — each one closing a gap in their own day, in a matter of hours.
Everyday tools rarely do exactly what we want. There is always a missing view, a missing calculation, a missing connection between two data sources. And even with AI, a conversation stays a conversation: you re-explain the context, rephrase the request, interpret the results. For recurring, specific needs, that is far from ideal. Build with Forra solves both problems at once — applications with a custom interface, connected to the tools you already use, with no technical background required.
At Mirego we do not just develop Forra, we use it every day. For several months now, teams with very different profiles have been building their own tools. Here are a few of them.
AI-powered deep research
When researching a complex topic, AI chat tools hit a limit: they lean on their training data or pull from a handful of surface-level sources. Work that calls for genuine depth — multiple sources, a critical eye on how reliable each one is — needs something else.
Dereck Bélanger, developer, built Deep Research in Forra. The application gathers information across many sources on the web, produces a structured and detailed document, assigns an authority score to each source, and flags the biases and gaps it finds. It runs asynchronously: launch a search, close the application, come back later for the results.
Building it required the SDK, and therefore specific technical skills. The result speaks for itself. Where a search used to return one or two surface-level results, Deep Research produces multi-page analyses backed by around ten sources. Dozens of people at Mirego have adopted it since.
Monitoring the performance of 16 digital products at a glance
Visualizing data in Google Analytics is a chore. Data Studio makes it better, but still tedious. Monitoring every product Mirego builds for its clients multiplies the properties, the filters and the reports. Simon Dostie, Director, Research and Data Intelligence, spent a good hour a week on it when time allowed.
He deployed SessionStart in Forra in about an hour. The application calls the Google Analytics properties directly, generates the visualizations, and writes a generative AI analysis under each chart. The result is a dashboard that consolidates traffic from 16 products into one view and produces an executive summary, validated by the research team before going out to account directors every Friday, with real-time alerts where they matter.
Tracking sprint progress by objective
In agile project management, a sprint usually holds several objectives, each made of many tasks. Knowing where each objective stands mid-sprint takes tracking Jira does not offer natively. Rémi Mongeau, project manager, wanted to give his team a simple way to see their progress without digging through the tool.
He built Sprint Objective on top of the Jira skill. The application pulls tasks, statuses and story points straight from Jira to calculate the real progress of each sprint objective. He iterated on the calculation, weighting a percentage against each task’s status and estimated effort.
It is now shared with the whole development team. They get real-time, self-serve visibility on progress by objective, which helps them make the right micro-decisions without losing sight of the sprint targets.
Reconciling a living backlog with initial estimates
Initial estimates live in a document, but the real backlog moves constantly: tasks get refined, split and added week after week. Knowing whether you are still inside the budget takes tracking nobody has time to do by hand. Rémi Mongeau had cobbled something together in a Google Sheet, but keeping it in sync with what was actually in Jira was never easy.
He built Découpage in Forra to make that link automatically. Using the Jira skill1, the application pulls tasks in real time and matches them against the estimates in the original document, letting AI establish the connections between the two. Rémi iterated for roughly two hours, with targeted adjustments each pass.
The application now gives the project manager and the product owner a clear, current view of where the project stands against what was planned — and a better footing for decisions with the client as it moves forward.
A command centre for project teams
Project teams at Mirego work in a deliberately rich, but fragmented, ecosystem. Jira, GitHub, Figma, Harvest, Slack, Notion: each tool lives in its own tab, every new mandate means hunting down the right workspaces, and every context switch costs attention. Generic AI assistants do not help — they see neither the open page nor the current mandate, so you end up exporting text, attaching screenshots and re-explaining the work with every question.
Marc Barry, QA specialist, built Outpost to solve both problems at once. Built directly on the Forra SDK2, which lets external applications connect to the platform and tap its AI capabilities, Outpost is a macOS application that brings web platforms, AI assistant, terminal and code into a single window organized by project. Each project keeps its own tabs, integrations and AI conversation. Page Insight reads what is on screen and proposes contextualized actions — summarize a page, write up a bug, draft a user story — without leaving the flow.
Outpost is concrete proof that Forra can power production applications well beyond the conversational interface. The full case study covers how it was built.
Centralizing the creation, tracking and performance of campaign links
Creating UTM tracking links, shortening them, keeping a history, following their performance: in marketing these tasks usually span several separate tools and a lot of back and forth. Marie-Septembre Larouche, marketing specialist, built UTM Builder in Forra to bring them into one application with three views.
The Generator offers a structured form for creating UTM links with the right parameters — source, medium, campaign, term, content — with no room for formatting errors. Every generated link is shortened automatically through Bitly, connected to the application via Bitly’s MCP server3. The second view keeps the full history of created links with their parameters and shortened versions. The third pulls campaign performance straight from Google Analytics, through the skill the Forra team developed.
What used to mean juggling a UTM generator, Bitly and Google Analytics now happens in one interface, built in a matter of minutes with Build with Forra.
Three ways to reinvent time tracking with Harvest
Harvest, the time-tracking tool used at Mirego, sits at the centre of the team’s day, but every role has needs it does not cover natively. Rather than wait for a feature or improvise in spreadsheets, three team members built their own applications in Forra.
All three were made possible by a Harvest skill1 the Forra team developed upfront. It lets applications read time entries by person or by project, generate summaries by period, aggregate hours by role, list the tasks assigned to a project, and create new entries — directly from Forra. That foundation is what lets non-developers build applications connected to Harvest without writing a line of integration code, while respecting Mirego’s security and governance requirements.
A dashboard that makes project management easier
Christian Dubois, project manager, needed to track a development budget spread across many parallel Harvest projects. Existing tools made a consolidated picture hard to draw. His application pulls it all into one dashboard with a real-time view of budget consumed, and the ability to drill into the detail to see what is eating hours or where there is still room. Built in a matter of minutes, it is now used daily by the project manager, the product owner and the business analyst.
A rethought interface for time entry
Marc-Olivier Fiset, developer, wanted to log hours in a calendar-style interface with start and end times, rather than entering duration blocks and working out each day how long every task took. His Better Timesheets application does exactly that: create blocks visually, time is calculated automatically, everything syncs to Harvest. Total build time, about three hours.
A tool that simplifies weekly team planning
Simon Dostie, Director, Research and Data Intelligence, needed a clear picture every Monday of what his team had worked on the week before. Getting that view in Harvest took enough manual filtering that the prep was happening on Sunday nights. His Time Validation application plugs straight into Harvest, breaks the information down by initiative — client work, business development, conferences — and generates an AI analysis. The Sunday night prep is gone.
A product built by the people who use it
What stands out across these examples is the diversity. Diversity of roles, of problems, of solutions built. But there is one common thread: each person identified a friction point in their own day and solved it themselves, in a matter of hours, without waiting for a third-party tool to do it for them. It is a natural extension of how Mirego approaches AI at work — not as a replacement, but as a lever that lets each person solve their own problems, in their own words, at their own pace.
It is also what makes Forra better. A digital product gains depth when the team building it uses it every day. Every feature is tested in real conditions before it ships. Every friction point is felt from the inside. The Harvest and Jira skills do not exist because they were on a roadmap; they exist because team members needed them to do their work. Build with Forra came from the same logic, and it is now available to every Forra user.
Want to see what your teams could build with Forra?
1 Skills in Forra give assistants the ability to run functions deterministically, particularly when calling external tools such as Jira, Harvest or Google Analytics. Rather than letting the AI interpret an API its own way, the code behind each skill guarantees that data is retrieved and structured correctly every time, and that only the functionality the company permits is reachable. Forra also supports the MCP protocol, and lets non-technical users build their own skills with Build with Forra.
2 The Forra SDK lets development teams connect their internal or external applications directly to the platform, to integrate AI capabilities such as assistants, skills and model streaming.
3 An MCP server (Model Context Protocol) is a standardized interface, based on an open protocol, that lets an AI application connect to external tools and services. In UTM Builder’s case it is an MCP server developed by Bitly, which lets the application create and shorten links directly from Forra with no custom integration.