Burrak

Engineering role

AI Software Architect

An AI software architect that weighs the options against your code and team, then writes down the decision — and what you give up for it.

Bring it a decision that will be expensive to reverse — split the monolith or not, events or direct calls, build the module or buy the service — and it does the architecture work: reads how the system is divided today, names the business boundaries the code should follow, sets out two or three real options with their costs, and recommends one. The result is an architecture decision record your team can argue with, not a diagram that looks good on a slide.

Things you could ask for

Written the way you would actually say them. The agent plans the steps itself.

  • “We are four engineers and the monolith is getting hard to change. Read the repository and tell me whether to split it, modularise it in place, or leave it — and what would change your answer.”
  • “Map the business domains in this codebase — orders, billing, accounts, notifications — and show me where the code crosses those boundaries when it should not.”
  • “Write an ADR for moving order notifications from direct API calls to an event queue: context, three options, the decision, and what becomes harder.”
  • “Plan how we replace the old reporting module without a rewrite: the order to move things in, what runs side by side, and how we know each step worked.”
  • “Read our last twelve decision records in Confluence and tell me which ones the code no longer follows.”

What you get back

Architecture decision records with context, options, decision and consequences; a map of domains and the dependencies that cross them; trade-off tables that name what each option costs; step-by-step evolution plans that avoid a big-bang rewrite; and architecture notes written into Confluence, Notion or Google Docs, with the files behind each claim.

Tools it works with

A role is a job description; connectors are what let it do the job on your real work instead of on whatever you paste into a chat.

GitHubGive an agent your repositories — issues, pull requests and the code itself.GitLabPut an agent to work on merge requests, issues and CI failures.Atlassian (Jira + Confluence)One connector for Jira and Confluence — the tickets and the pages that explain them.NotionGive an agent your Notion workspace as working memory.Google DocsAn agent that drafts and edits inside the document your team already comments on.LinearLet an agent triage, plan and update Linear issues on its own.

Where it is the wrong tool

  • It recommends; your team decides. An architecture is only as good as the team’s willingness to maintain it, so treat its ADR as the start of the discussion, not the end of it.
  • It reads code, not running systems. A remote agent has no shell in the cloud and no cloud-account connector, so latency, load and what is actually deployed are unknowns it will name rather than guess. Measure them, or run a local agent with shell access on your machine.
  • It does not do the migration. It plans the order of the change; the code is written and reviewed in your normal pull requests, and the GitHub connector cannot push or merge.
  • It works best with constraints. Without your team size, budget, deadlines and what you already run, any recommendation is generic — give it those first.

Starting an agent with this role

  1. 1Create a Burrak account — any plan works, and your first week is $1.
  2. 2In the Marketplace, open the Roles tab, find Software Architect and choose “Start an agent as Software Architect”. That creates a remote agent with the role’s instructions in its system prompt.
  3. 3Connect the tools it needs — GitHub, GitLab, Atlassian (Jira + Confluence), Notion, Google Docs, Linear — from Connectors.
  4. 4Give it a brief. It keeps working in the cloud after you close the tab, and you are only charged while it is actually working.
Start for $1How agents work

The Software Architect role, in practice

Should we move to microservices?
Usually not yet, and it will tell you why for your case: team size, how often the parts change independently, and what you are prepared to operate. When the answer is no, it says what would have to change for it to become yes, so the question does not come back every quarter.
How is this different from the Backend Architect?
The Backend Architect designs a specific change: the schema, the API contract, the migration and its rollback. The Software Architect works one level up — where the boundaries between parts of the system belong, which pattern fits, and the record of why. Decide the shape with the Software Architect, then hand the change to the Backend Architect.
Can it document an architecture that was never written down?
Yes. It reads the repository, maps the domains and dependencies as they are rather than as anyone remembers them, and writes the records for decisions that were made but never recorded — marking each one as reconstructed from the code.

In the docs

Projects & workspacesRead the docs →Teams & collaborationRead the docs →

Capability reference

Summarised from the role’s instructions, which are adapted from the agency-agents collection (AgentLand Contributors), used under the MIT licence.

  • Domain modelling: bounded contexts, aggregates and domain events
  • Pattern selection: modular monolith, microservices, event-driven, hexagonal
  • Trade-off analysis that names what each option gives up
  • Architecture decision records: context, options, decision, consequences
  • Evolution strategies that grow the system without a rewrite
Your first 7 days for $1

Your AI assistant is waiting.
Put it to work today.

Put an agent on the work that never needed a human in the first place — the research, the reports, the follow-ups — and take the week back.

Start for $1Download desktop appTry in browser
App StoreGoogle Play

$1 for your first week · Cancel anytime · Works on Mac, Windows, iPhone, Android & Web