Early access — free for one production

Your whole SDLC as one living knowledge graph

Spectackl replaces hundreds of drifting documents with a single connected model of your system — requirements, architecture, APIs, tickets and tests — that your team and your AI agents work on together.

Free forever for one production. No credit card required.

Acme Theatre / Payments Platform / Architecture

Trace — requirement to test

  1. REQ-14Requirementchanged

    Customer can pay with a stored card

  2. ARC-3Architecture

    Payments service — tokenised vault

  3. API-27APIin review

    POST /payments/charge

  4. TSK-118Task

    Implement charge handler

  5. TST-52Test

    Charge with expired token returns 402

Impact analysis

REQ-14 changed. 5 downstream artifacts affected, 1 breaking.

  • ARC-3ArchitectureStructural
  • API-27API contractBreaking
  • TSK-118TaskRework
  • TST-52TestUpdate
  • DOC-9Integration guideRegenerate

Architecture agent drafted an updated contract

Waiting for human approval

Plugs into the tools your team already works in

  • Model Context Protocol
  • Claude Code
  • VS Code
  • Cursor
  • GitHub
  • CLI & CI

The problem

Documents drift. A graph doesn’t.

Every serious system is described in hundreds of files that disagree with each other and with the code. Spectackl replaces them with one structured, versioned model.

Knowledge is scattered

Requirements in one tool, decisions in a wiki, diagrams in another, the truth in someone’s head. Nothing reconciles.

Traceability is a manual audit

Proving that a shipped change maps back to an approved requirement means a week of spreadsheet archaeology.

Impact analysis is guesswork

Change one contract and you find the consequences in production, not in review.

AI has no context

A chat window that cannot see your architecture writes plausible code against the wrong model.

Platform

One model, the entire lifecycle

Spectackl is not another document store with a chat box bolted on. It is a knowledge graph of your system that humans and agents edit under the same rules.

Unified artifact graph

Requirements, analysis, architecture, domain and data models, integrations, APIs, UI designs, tickets, tests, docs and governance — structured, versioned, explicitly linked.

End-to-end traceability

Every task, test and change resolves back to an approved requirement. Coverage is a query, not a quarterly exercise.

Automated impact analysis

Propose a change and see every dependent artifact and flagged risk before anyone writes code.

Specialised AI actors

Product, analyst, architect, developer and QA agents that operate on the graph — with your real architecture as context, not a pasted snippet.

Human-in-the-loop governance

Agents propose, people decide. Moderation, approvals, quality gates, RBAC and a full audit trail are part of the model.

Generated docs & diagrams

Specifications, API references and architecture diagrams rendered from the live graph, so documentation cannot go stale.

How it works

From scattered documents to a directed production

Spectackl borrows its language from the theatre: your organisation is a Theatre, each project a Production, work happens in Scenes, and people and agents are Actors.

01

Model your production

Point Spectackl at an existing repository and its documents, or start from a blank stage. Reverse engineering turns what you already have into artifacts and relations — you do not migrate everything by hand.

02

Direct the actors

Open a scene, put the spotlight on what matters, and assign work to people and agents side by side. Every proposal lands as a versioned artifact with its links already drawn.

03

Ship with the trail intact

Approvals, quality gates and impact analysis run against the graph. Docs, diagrams and coverage update themselves as the model moves.

Agent-native

Your agents get the whole system, not a snippet

Spectackl speaks the Model Context Protocol natively. Connect Claude Code, your IDE or your own agents and they operate on the same structured artifacts, relations and governance rules your team does.

  • MCP server — your coding agent reads and writes the graph directly, with tenant-scoped permissions.
  • CLI and CI hooks — create artifacts, claim tasks and gate builds from any pipeline.
  • Agent marketplace — publish or acquire agents, skills, prompts and workflows across organisations.
  • Model-agnostic — bring the assistants your team already pays for.
.mcp.json
{
  "mcpServers": {
    "spectackl": {
      "type": "http",
      "url": "https://mcp.spectackl.ai/mcp",
      "headers": {
        "Authorization": "Bearer ${SPECTACKL_API_KEY}"
      }
    }
  }
}
Terminal
$ spectackl artifact create \
    --kind REQUIREMENT \
    --title "Customer can pay with a stored card"

REQ-14 created · linked to ARC-3 · 2 agents notified

Pricing

Start free. Pay when your theatre fills up.

Every plan ships the complete artifact model — nothing about the graph itself is locked behind a bigger tier.

Personal

For solo builders and anyone evaluating Spectackl on a real system.

Free

Free forever · no credit card

Includes

  • One production (project)
  • Full artifact graph and traceability
  • Impact analysis on your production
  • MCP server and CLI access
  • Bring your own model keys
  • Community support
Most popular

Organization

For teams running several systems with people and agents side by side.

$39per seat / month

Billed annually · Agents do not consume seats

Everything in Personal, plus

  • Unlimited productions
  • Teams, roles and full RBAC
  • Governance workflows and approval gates
  • Cross-production impact analysis
  • Agent marketplace — publish and acquire
  • Audit trail and usage telemetry
  • Email support, next business day

Corporate

For regulated and security-led organisations that keep the platform inside their own perimeter.

$79per seat / month

Billed annually · Minimum 25 seats · annual agreement

Everything in Organization, plus

  • On-premise or your own cloud deployment
  • SSO — SAML / OIDC — and SCIM provisioning
  • Private model routing and data residency
  • Custom compliance and retention policies
  • Audit log export to your SIEM
  • 99.9% uptime SLA
  • Dedicated onboarding and support channel

Prices in USD, excluding local taxes. Annual plans are billed once per year. Ask about non-profit and academic pricing.

FAQ

Questions teams ask before they move

Is this yet another documentation tool?

No. Documents are the output, not the storage. Spectackl keeps a structured, versioned graph of requirements, architecture, APIs, tasks and tests, and renders documents and diagrams from it on demand — so the docs cannot drift from the model.

Do we have to migrate everything before we get value?

No. Start with one production and let reverse engineering derive artifacts from the repository and documents you already have. The graph grows as you work; nothing has to be re-typed up front.

What exactly is a production?

A production is a project — the whole effort to design, build and evolve one system, including its artifacts, actors, workflows and history. The Personal plan covers one; Organization and Corporate cover as many as you need.

Which AI models can we use?

Spectackl is model-agnostic. Connect the assistants your team already pays for through the MCP server, or bring your own provider keys. On Corporate you can route agent traffic to a private or self-hosted model.

Can agents change our system without review?

Only if you let them. Agents produce proposals as versioned artifacts; moderation, approval workflows and RBAC decide what is promoted. Every action lands in the audit trail with its actor.

How do seats work?

A seat is a human actor in your theatre. AI agents, service accounts and CI integrations do not consume seats — you are billed for the people, not the automation.

Can we run Spectackl in our own environment?

Yes, on the Corporate plan. Deploy on-premise or into your own cloud account, with SSO, SCIM provisioning and data residency under your control.

What happens to our data if we leave?

The graph is yours. Export every artifact, relation and version as structured files, or keep using them through the API and CLI during a wind-down.

Put your first production on the stage

Model one system, connect your agents, and see the trace from requirement to test in an afternoon.