Specsight AI product observability for teams is a simple idea aimed at a very modern pain: your product now changes faster than anyone outside engineering can follow. When AI coding tools let a small team ship several releases a week, release notes go stale, documentation lags behind, sales demos describe last month’s product, and support answers questions about features that no longer work the way the help centre says they do. Specsight, launched in April 2026 by founder Ola Piętka, tries to close that gap by reading the source code itself and turning it into a living picture of the product in plain language. In this review, we look at what it actually does, what it costs, who it suits, and where its limits lie, so you can judge whether it belongs in your stack.
What Specsight does
Specsight connects to a GitHub or GitLab repository and branch, then analyses the code to build what it calls a product map: what features exist, how they behave, how they connect, and what changed with every release. The company says most projects get their first product map within about an hour. From then on, each release updates that map automatically, so the description of the product is derived from what actually shipped rather than from a document somebody remembered to update. Its Product Hunt launch page and its official site describe the same promise in the founder’s own words.
That distinction matters. Traditional documentation is written about a product; Specsight’s product spec is generated from the product. It describes features and behaviours using Context, Action and Outcome scenarios, which is a useful discipline: instead of a vague feature name, you get the situation in which it applies, what the user does, and what the system does in return. For a support agent answering a ticket, or a salesperson preparing a demo, that level of detail is the difference between a confident answer and a promise to check with engineering.
Specsight AI product observability for teams: the five pieces
The product is organised around five linked pieces, and it helps to understand them as one workflow rather than five separate features.
- Product changelog. Every release produces a clear picture of what changed and how it affects customers, in user-facing terms. The example the company gives is telling: rather than reporting a code change, Specsight should report that a file-upload limit rose from 10 MB to 100 MB.
- Product map. You can follow any user journey from start to finish and see how features connect, without clicking through the live application to reconstruct the flow yourself.
- Product spec. Features and behaviours are documented from the actual codebase and kept up to date as the code changes, so the spec cannot quietly drift from reality.
- Product answers. Anyone can ask a question about the product and get an answer in seconds, inside Specsight or through MCP in the AI tools they already use.
- Product updates. Weekly team digests arrive automatically, and release notes are drafted for you, which removes the most commonly skipped writing job in a fast-shipping team.
There is also a live demo at specsight.app/demo where you can see Specsight tracking itself. That is a sensible thing to try before buying: if the demo does not make your own product’s complexity clearer, the tool will not magically do so for a messier codebase.
Pricing: one standard plan, plus custom
Specsight’s published pricing is refreshingly short, which is both a strength and a warning sign. There are two tiers on its site at the time of writing. The Standard plan costs €150 per month and includes all Specsight features, up to three projects, up to 50 members, and 200 syncs a month. The Custom plan is priced on conversation, for bigger teams, more projects, or teams releasing every day.
Do the arithmetic before you sign up. Two hundred syncs a month is roughly six to seven a day. That is plenty for a team shipping a few releases a week, but a team deploying many times a day across three projects could hit the ceiling quickly, at which point the real price is the Custom plan, not €150. Similarly, 50 members sounds generous until you count sales, support, product and marketing across a 200-person company, which is the top end of Specsight’s stated target market of B2B software companies with 20 to 200 employees in Europe and the US. The pricing page does not advertise a free tier, so budget for a paid trial period and judge it on a real release cycle, not on the demo.
For Product Hunt visitors, Specsight offered 50% off the first three months with code PH2026 until 8 November 2026. If you are reading this after that date, treat the offer as expired and judge the full price on its own merits.
Who Specsight is for
The clearest fit is a B2B software team shipping frequently with AI assistance, where the bottleneck has moved from writing code to explaining what the code now does. That is exactly the shift the founder describes: building is fast now, keeping up with what got built is not. If your product manager spends half the week translating engineering output for the rest of the company, if sales keeps demoing features that changed last sprint, or if support learns about changes from angry customers, Specsight addresses a real, expensive problem.
It is less obviously useful for a team that ships rarely, has a small and stable feature set, or already maintains excellent documentation discipline. A product that changes once a quarter does not need a living product map; a well-kept changelog and a quarterly docs review will do the same job for free. Likewise, a very early startup with one product, one repository and five people in one room probably does not need a €150-a-month translation layer. The value grows with the number of people who need to understand the product but do not read code, and with the speed at which the product changes.
Watch-outs and honest limitations
First, Specsight is young. It launched in April 2026, and independent reporting in late October 2026 described five closed deals and two new pilots since June, with some early customers cancelling. That is normal for a young product, but it means you should not buy it as if it were a mature, battle-tested platform. Ask for references from companies shaped like yours, and test it on your own repository during a real sprint before committing the wider team.
Second, the quality of the output depends on the quality of the input. A codebase with clear feature boundaries, sensible naming and well-structured flows will produce a clean product map. A tangled codebase full of dead code, experimental branches and half-finished features will produce a map of that tangle, faithfully. Specsight translates code into plain language; it does not tidy the underlying product. If your map looks confusing, that may be the most valuable thing it tells you, but it is not the outcome the marketing promises.
Third, access and privacy deserve a direct question. Specsight says it uses read-only access to your repositories, and that code analysed through Anthropic’s Claude API is deleted afterwards rather than stored. That is a good starting answer, but read-only still means a third party reads your entire codebase, and teams in regulated industries or with strict customer contracts should confirm the data-processing terms, retention details and any sub-processors in writing before connecting a production repository. Start with a single, non-sensitive repository if you have any doubt.
Fourth, the company itself describes the current product as an observability phase. Its longer-term ambition is to become a control centre where teams verify changes made by coding agents and assess their business impact, but monitoring and broader control capabilities remain on the roadmap. Buy Specsight for what it does today — tracking and explaining what shipped — not for a control-centre vision that has not shipped yet.
How it compares to alternatives
The honest comparison is not with another single tool, because Specsight sits across several jobs that teams currently do by hand. Release notes are usually written in Notion, Confluence or a changelog tool; product documentation lives in a wiki; support answers come from tribal knowledge; sales knowledge comes from shadowing engineering. Each of those is cheaper individually, and each goes stale independently. Specsight’s pitch is that deriving all of them from the same source — the code — keeps them consistent with each other.
From our experience: who should and should not buy
From our experience building and shipping software for clients over nine years and more than 300 projects, the pattern is consistent: the teams that suffer most from product drift are not the ones that ship badly, but the ones that ship fastest. AI coding tools have made that worse, because a feature can now go from idea to production in a sprint without ever passing through a product conversation. The non-technical teams are not lazy; they are simply outpaced.
Our practical advice is to run a two-sprint test. Connect one repository, let Specsight build the map, and then measure three things: how often support still has to ask engineering a question it should have been able to answer itself, whether sales demos stop containing outdated claims, and whether the weekly digest actually gets read. If all three move, the tool pays for itself quickly, because a lost deal or a churned customer over a slow, wrong answer costs far more than €150. If none of them move, the problem is probably not observability; it is product complexity, unclear ownership, or a release process that needs fixing first, and no tool will fix that.
Do not buy it as a documentation replacement on day one. Keep your existing docs running in parallel for the first month, compare the two, and only retire the manual process once Specsight has survived a messy release — a reverted feature, a renamed flow, a breaking change — without misleading anyone. That is the test that matters.
Frequently asked questions
What is Specsight AI product observability for teams?
It is a platform that reads your GitHub or GitLab codebase and turns it into a living product map, changelog, product spec and question-answering layer, all in plain language, so product, sales and support can keep up with fast AI-assisted shipping without chasing engineering.
How much does Specsight cost?
The Standard plan is €150 per month for all features, up to three projects, up to 50 members and 200 syncs a month. Larger teams or higher release volumes need the Custom plan, priced on conversation. Check the pricing page before buying, as early-stage pricing can change.
Does Specsight write code or change my product?
No. It uses read-only access to your repositories and observes what shipped; it does not modify code. Its current phase is observability — understanding what changed — with verification and control features described as future roadmap items.
Is my code stored by Specsight?
Specsight says code analysed through Anthropic’s Claude API is deleted afterwards rather than stored, and that repository access is read-only. Confirm the current data-processing terms in writing before connecting a sensitive repository, especially under customer or regulatory constraints.
Should a small team buy Specsight?
Probably not until product drift is actually costing you time or deals. A team that ships rarely, or where everyone already reads code, will get little value. The fit improves sharply once several non-technical teams depend on understanding a fast-changing product.
Specsight is a well-aimed answer to a genuinely new problem: AI lets teams build faster than their companies can understand what they built. At €150 a month it is cheap to test. Try it on one repository over two sprints, measure support questions, demo accuracy and digest readership, and only then roll it out. If your team needs help building the surrounding discipline — release processes, documentation workflows or AI-assisted development done properly — talk to MTD Technologies or get in touch, and explore more tools in our AI Tools Directory.