Guides  /  Rollout and delivery

Where does spec-driven delivery fit with the tools we already use?

Answer

Akka Specify runs inside the repositories, CI systems and AI coding assistants a team already uses. The specification is versioned beside the code it describes, and the check runs where the change is written.

The same check reaches a change in the editor and again in CI. A developer sees a refusal while writing, and the branch is checked once more before it merges.

The mechanism

A separate review step runs after the change is written.

Governance that lives in its own tool depends on someone opening it. A team writes the change first and reconciles it with the document afterwards, and the reconciliation happens on the schedule of whoever owns the document.

The Akka Specify check runs in the places changes already pass through. A repository holds the specification beside the code. CI runs the check on the branch. An AI coding assistant reads the specification while it writes.

Where the check runs

The check runs in the editor, in CI, and at build time.

The repositoryThe specification is versioned beside the code, so a change to a requirement carries an author, a diff and a review.
The AI coding assistantThe assistant reads the specification as it writes, so a generated change starts inside the requirement.
Continuous integrationThe branch is checked against the specification before it merges, and a change that contradicts it is refused.
The build and deploy gateAn evaluation suite runs at build time and fails the deploy when the signed control set is unmet, producing a versioned, hash-chained report.
What stays where it is

The existing stack keeps its work.

The major model, observability, relational and graph database, messaging and evaluation vendors are integrated into the Akka runtime, so a team composes with the AI stack it already runs. The platform also integrates with enterprise IAM and with the golden-path workflows a platform team already operates.

Observability tools stay useful for engineering work such as latency graphs and model A/B testing. Governance evidence comes from the runtime record, because a tool reading logs from outside the request path samples by design.

Questions

Related questions.

Do we have to move our repositories?

The specification is committed into the repository the code already lives in, and Akka Specify reads it there.

Which AI coding assistants does this cover?

Akka Optimize grades work produced in third-party harnesses including Cursor, Claude Code and Copilot. Akka Specify supplies the specification an assistant writes against.

What happens to our identity and access system?

The platform integrates with enterprise IAM, so the identity a developer already holds is the identity the platform recognises.

Do we have to stop using our current agent framework?

Akka Verify evaluates agents built elsewhere and keeps the evidence record for them. Inline enforcement requires the runtime to sit in the execution path, which means the agent runs on Akka.

Where does the specification live for a system Akka delivered?

The specifications are handed over with the running system and maintained by the customer. Extending the system means editing the specification the check reads.

On Akka. Specification-checked delivery is Akka Specify. Evaluation and evidence for agents built elsewhere are capabilities of Akka Verify. Grading against third-party harnesses is part of Akka Optimize.