StackWeavers
Start free
Part of the platform

Hire a specialized agent.

Pay only for what they ship. Pair it with a human when it matters.

The default StackWeavers runtime ships full vertical slices end to end. Specialized Agents are scoped to one discipline at a time. Development, architecture, code review, security, migration, documentation. Use them when the work is a focused problem instead of a complete feature, and pair any of them with a human reviewer for the high-stakes call. Same quality gates. Same no-lock-in. Billed in Capacity Units per delivered unit.

Book a demoSee the roster

section/01 · roster

The roster.

Six specialist roles. Each one is scoped to a single discipline and ships a focused deliverable. Same quality gates as the default runtime. Same lineage tracking. Same BYOK or managed billing. Pair any of them with a human reviewer.

Development Agent

A focused engineer for one task at a time.

Implements a single feature task with full test coverage and contract-conformant code. Scoped to one ticket. Ships a clean pull request that goes through your normal review process.

Delivers
Pull request with implementation, tests, and contract docs.
Best for
Teams that want help on one ticket without the full vertical-slice flow.
Indicative cost
50–200 CU per task
See it in a demo →

Architect Agent

System design, RFCs, and migration plans.

Drafts the design before any code gets written. Compares technology choices with tradeoffs. Writes RFCs and architecture decision records. Plans multi-step migrations across services and repos.

Delivers
RFC, design document, ADR, or migration plan with options and tradeoffs.
Best for
Pre-build planning, technology selection, and migration scoping.
Indicative cost
200–600 CU per RFC
See it in a demo →

Lead Reviewer Agent

Senior-engineer review on top of the gates.

A second opinion on high-stakes pull requests, layered on top of the six automated quality gates. Reviews through an architectural lens: contract drift, abstraction quality, test coverage of intent, code reuse, and the avoidance of duplicates.

Delivers
Pull request review with inline comments, approve or request-changes, and a rollup summary.
Best for
Changes where the automated gates aren't enough. Sensitive code paths. Pre-release.
Indicative cost
30–80 CU per PR
See it in a demo →

Security Auditor Agent

Threat models, dependency audits, secrets sweeps.

Audits a vertical slice or a defined surface area for security issues. Threat models the data flow, audits the dependency tree, sweeps for committed secrets, and reviews the auth surface.

Delivers
Security audit report with findings, severity, and recommended remediations.
Best for
Pre-release reviews, post-pen-test follow-ups, compliance preparation.
Indicative cost
150–400 CU per audit
See it in a demo →

Migration Specialist Agent

Framework upgrades and deprecation runways.

Handles framework upgrades, language version bumps, ORM migrations, and deprecation cleanup. Moves the codebase forward in incremental steps that don't break production. Each step ships as its own pull request.

Delivers
A series of incremental pull requests with green tests at every step.
Best for
Legacy modernization, framework upgrades, and post-deprecation cleanup.
Indicative cost
100–500 CU per step
See it in a demo →

Documentation Agent

Living docs that follow the code.

Writes and maintains documentation from the codebase. API docs, README files, ADR catalogs, and runbooks. Watches for code changes that drift from the docs and proposes updates as pull requests.

Delivers
Living documentation that stays in sync with the code through the platform's quality gates.
Best for
Teams that have never had time to write docs but now need them for hiring, audits, or handoff.
Indicative cost
50–150 CU per docset
See it in a demo →

section/02 · when-to-use

When to use Specialized Agents vs the default runtime.

The default runtime ships complete features. Specialized Agents ship a focused deliverable. Pick the right tool for what you actually need.

default runtime

Use the default runtime when you need a feature shipped end to end.

  • Full vertical slice. Database, API, frontend, tests, deploy.
  • Scope locked by a product owner before agents start.
  • Best for: net-new features and complete deliverables.

specialized agents

Use a Specialized Agent when the work is a focused expert problem.

  • One discipline at a time. RFC, code review, security audit, migration step.
  • Pair with the human HITL rate card for the high-stakes calls.
  • Best for: focused problems, second opinions, scoped reviews.

section/03 · pairs-with-hitl

Agents do the work. Humans review the high-stakes calls.

Specialized Agents pair naturally with the human-in-the-loop rate card. Use them together for the changes that matter.

Architect Agent + human Architect

The agent drafts three RFC options. A human architect picks the one that matches constraints the agent doesn't see (politics, hiring pipeline, regional cloud spend).

Lead Reviewer Agent + Senior Engineer

The agent reviews the diff line by line. The human reviewer signs off on the architectural intent and the impact on the team's working model.

Security Auditor Agent + Security & Compliance

The agent runs the full sweep. A human signs the audit report, owns the remediation roadmap, and presents to the auditor or board.

See the human rate card on /pricing.

About Specialized Agents

How is this different from the default StackWeavers runtime?
The default runtime ships full vertical slices end to end: database, API, frontend, tests, deploy. Specialized Agents are scoped to a single discipline. Use them when the work is a focused problem (an RFC, a code review, a migration step) rather than a full feature. Same quality gates apply either way.
Which plans include Specialized Agents?
Pro, Team, and Enterprise. The Free tier stays scoped to one project on the default vertical-slice runtime. On Team and Enterprise, you can pair an agent with included human-in-the-loop hours for the high-stakes calls.
How does pricing work?
Specialized Agents are billed in Capacity Units per delivered unit (PR, RFC, audit, migration step, or docset). The ranges on this page are indicative. They draw from your existing CU bundle on the managed track, or run free of CU caps on the BYOK track since you pay your model provider directly.
What makes this different from a pure agent swarm?
Every Specialized Agent pairs with an optional human reviewer from the rate card. The agent does the work; a named human signs off on the high-stakes call. That agent-plus-human model is the point. It is not a fully autonomous swarm you have to trust blind, and it is not a black box. You see every step, and a human owns the decision that matters.
Does it work with BYOK?
Yes. Specialized Agents run on the same model-routing layer as the default runtime. On managed, the platform handles model spend. On BYOK, you pay your model provider directly and the platform charges a flat fee, with no CU cap or per-token markup.
Can I bring my own specialized agent?
Not at this time. StackWeavers is a closed-source platform and we don't operate a third-party agent marketplace. The roster on this page is the catalog we ship and maintain ourselves.
How is this different from the human HITL rate card?
Specialized Agents are AI specialists. The HITL rate card buys you human specialists (Product Owner, Senior Engineer, Architect, Security & Compliance). You can mix the two: an agent does the work and a human reviews the high-stakes call. That pairing is the recommended way to run them.
How do I get started?
Book a demo and we'll walk through the roles that fit your work, or start on Pro and add a specialist to your first deliverable. The code stays in your cloud, your keys, with no training on it, the same as the rest of the platform.

Put a specialist on it.

Start building freeBook a demo