Mantrana — deliberation before code

Describe it rough. Ship it right.

Four specialist AI jurors challenge your idea and file one implementation plan, with unresolved issues called out.

For coding agents: Mantrana now speaks MCP — connect Claude Code →

What are you trying to build? A few honest sentences are enough.

Precedents

Specification depth

SPECIMEN · FILED RECORD № 01

AChief Systems Architect

Data contract boundaries resolved; idempotency keyed on webhook event ID.

RAdversarial Red Teamer

Replay vector on retry storm: debounce claim window to 30s.

PProduct & UX Strategist

First-run success signal: user copies spec within 90s of landing.

OImplementation & Operations

Ship order: contract tests before the dashboard UI. Gate: green CI.

A R P O

The Council

Four perspectives. One accountable brief.

Each juror examines architecture, risk, user value, or implementation. A separate quality review checks the plan and calls out unresolved issues.

Chief Systems Architect

System boundaries, data contracts, concurrency, and capacity.

Adversarial Red Teamer

Failure modes, security surfaces, races, and missed states.

Product & UX Strategist

User journeys, MVP scope, success signals, and friction.

Implementation & Operations

Delivery sequence, testing, telemetry, and release gates.

Inside your coding agent

Mantrana now speaks MCP.

Connect Claude Code — or any MCP client — and the council joins your agent's tool loop. Your agent files the brief with real evidence from your codebase, the same four jurors deliberate, and the reviewed plan lands in the session before code gets written.

Issue an API key

Sign in, open the account menu → Developer & credits, and create a key. Keys look like mk_…, are shown once, and revoke instantly.

Connect in one line

claude mcp add --transport http mantrana \
  https://mantrana.co/mcp \
  --header "Authorization: Bearer mk_YOUR_KEY"

The council joins the loop

Your agent calls mantrana_specify with the brief. The jurors deliberate — and can pause mid-run to ask you questions through your agent — then file the plan with its quality report straight into the session.

One wallet, every surface. A single USD credit balance funds plans on the website and through your agent — top up once, spend anywhere. Failed runs refund automatically.

Read the full MCP guide — tools, setup for every client, and examples →

Planning examples

See what a planning record covers.

These showcase examples illustrate how a specification can guide a build. They do not validate the current planning pipeline or predict your results.

Review the assumptions, implementation decisions, and verification steps before building.

Review depth

Choose the depth your build deserves.

Every tier includes plan synthesis and quality review. Plans with unresolved blocking issues are delivered as drafts and use normal run billing.

$10 signup credits, once per account. Use any tier on the website or through MCP. No card needed.

MVP Spec

$1.50

Intro launch price — 50% off the $3 list price for a limited time.

one specification

A lean, buildable brief for one sharply scoped feature.

Four independent specialist reviews
A plan shaped to your task and constraints
Quality review and any unresolved issues

Recommended

Full-Stack

$2.50

Intro launch price — 50% off the $5 list price for a limited time.

one specification

The complete implementation brief for a production feature.

Everything in MVP, plus:

Architect and Red Team cross-review
Review of assumptions and failure modes
Full critiques considered in plan synthesis

Enterprise

III

$5.00

Intro launch price — 50% off the $10 list price for a limited time.

one specification

A deeper systems brief for consequential, multi-team work.

Everything in Full-Stack, plus:

Product and Operations follow-up
Scope and feasibility checked against critiques
Unresolved objections carried into synthesis

Runs still queued after 30 minutes are canceled and refunded to your wallet.

Questions about adding credits? Contact admin@autogc.in.

Before you begin

A few practical answers.

What do I receive?

A Markdown specification you can copy or download, the specialist reviews, and a quality report. The plan sets out implementation decisions, verification steps, and unresolved issues for you or your coding agent to review.

How long does a run take?

Allow several minutes; timing varies with review depth, queue load, and the complexity of your brief. Progress appears as the council works. Runs can pause for your answers, which adds time. A run still queued after 30 minutes is canceled and its credits return to your wallet.

Am I charged for a draft?

Yes. A completed plan with unresolved blocking issues is labeled “Draft — review required” and uses normal run billing. Failed runs and runs canceled while queued receive a credit refund.

How is my data handled?

Your brief and supplied context are processed by AI providers to prepare the plan. Runs are saved to your account. Publishing creates a public share of the plan and review text; raw planning context and private repository evidence packs are excluded, but quotations may appear in the output. Review the content before sharing and keep credentials out of your brief.