Engineering role
AI Backend Architect
An AI backend architect that reads your code and your schema, then designs the change — with the migration, the API contract and the failure modes written down.
Give it your repository, your database and a problem — a slow endpoint, a table that has to change shape, a service that keeps falling over — and it does the design work: reads how the system is built today, proposes the simplest change that solves it, and writes out the schema, the API contract, the migration plan and what happens when each dependency fails. It reaches for a monolith before microservices and says why.
Things you could ask for
Written the way you would actually say them. The agent plans the steps itself.
- “Read our repository and write a one-page description of the backend as it actually is: services, data stores, how requests flow, and the three riskiest parts.”
- “We need to split the users table into accounts and profiles without downtime. Write the expand-and-contract migration plan, step by step, with the rollback for each step.”
- “Design the API for our new webhooks feature: endpoints, payloads, retries, idempotency keys and error format, as an OpenAPI spec.”
- “This endpoint is slow. Read the code and the query, run EXPLAIN on a Neon branch, and tell me what index or rewrite fixes it.”
- “Go through last week’s Sentry errors and tell me which ones point at a design problem rather than a bug.”
What you get back
Design documents you can review in a pull request: architecture descriptions, schema changes as SQL, API contracts as OpenAPI, migration plans with rollback steps, and decision records that state the trade-off taken. Code comes back as files or diffs; query work run on a Neon branch comes back with the plan and timings.
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.
Where it is the wrong tool
- It does not deploy. Designs and diffs come back to you for review; what gets merged and released is decided by your own review process and pipeline, not by the agent.
- A remote agent cannot run your service or your tests — it has no shell in the cloud. It reasons from the code, and load or latency claims need a measurement on your side. A local agent with shell access in the desktop app can run them on your machine.
- Keep it off production writes. The Neon and Supabase connectors can run SQL, including statements you would not want run twice; give it a branch or a development database.
- It has no cloud-account connector. It can review Terraform or Kubernetes files in your repository, but it cannot see what is actually running in AWS, Google Cloud or Azure.
Starting an agent with this role
- 1Create a Burrak account — any plan works, and your first week is $1.
- 2In the Marketplace, open the Roles tab, find Backend Architect and choose “Start an agent as Backend Architect”. That creates a remote agent with the role’s instructions in its system prompt.
- 3Connect the tools it needs — GitHub, GitLab, Neon, Supabase, Sentry, Grafana — from Connectors.
- 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.
The Backend Architect role, in practice
- Can it design a system from scratch?
- Yes, and it is most useful when it has your constraints: team size, expected load, what you already run. It will start with the simplest architecture that meets them and write down when you would need the next step, rather than draw a microservices diagram for a team of three.
- How is this different from the Code Reviewer?
- The Code Reviewer reads a pull request after it is written. The Backend Architect works before the code: the schema, the API, the migration order and the failure modes. Use the architect for the design and the reviewer on the pull request that implements it.
- Will it see my production data?
- Only if you connect a production database, which we advise against. Connect a Neon branch, a development database or a read replica; schema, query plans and a sample of rows are what design work needs.
Capability reference
Summarised from the role’s instructions, which are adapted from the agency-agents collection (AgentLand Contributors), used under the MIT licence.
- Chooses monolith, modular monolith, microservices or serverless from team size and real scaling needs
- Zero-downtime schema migrations using expand-and-contract, with backfills and rollback planned first
- API contracts in OpenAPI or equivalent, with versioning, idempotency keys and a standard error format
- Timeouts, retries with backoff, circuit breakers and dead-letter queues for every external call
- Observability by design: structured logs, service-level objectives and alerts on user-facing symptoms
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.
$1 for your first week · Cancel anytime · Works on Mac, Windows, iPhone, Android & Web