Design an AI agent pipeline or multi-agent orchestration system.

USE THIS WHEN:

• You are building a system where LLMs call tools or hand off to other agents

• You want to visualise a multi-agent workflow (planner → researcher → writer)

• You need to document tool access, memory strategy, and failure behaviour

NOT FOR:

• General system architecture → use umlforge_stakeholder_arch

• Standard synchronous API flows → use umlforge_api_sequence

• Event-driven pipelines without LLM agents → use umlforge_event_driven

Produces:

• Agent pipeline sequence: agents as participants with model names, tool calls as self-calls, human-in-the-loop gates, retry/fallback logic

• Agent component map: agents, tool deps, memory, external integrations

• Agent responsibility matrix: model, role, tools, inputs, outputs, failure behaviour

• Risk & observability note: hallucination hotspots, validation gates, logging points

• Tool overload flags (⚠️) for agents with more than 5 tools

• (report_mode=True) Agent Pipeline Analysis Notes: pipeline risks, coverage gaps, reliability quick wins

Args: pipeline_purpose: What the agent system does (e.g. "Research pipeline that queries the web and drafts a report"). agents: Agents and their roles (e.g. "Planner [claude-opus-4], Researcher [claude-sonnet-4]"). tools_available: Tools agents can call (e.g. "web_search, execute_code") (optional). orchestration_approach: Coordination strategy — sequential, DAG, hierarchical, parallel fan-out (optional). memory_strategy: Memory approach — shared context, vector memory, Redis, none (optional). report_mode: True → also produce Agent Pipeline Analysis Notes. Pro/Team/Enterprise only. Omit to use your saved account preference (dashboard toggle); False forces diagrams only.

Diagram how services call each other for a specific user action or API flow.

USE THIS WHEN:

• You want to trace a request across multiple services (e.g. "user logs in")

• You are designing API contracts between microservices

• You need to show failure paths, retries, and error handling between services

• You are writing QA test cases or doing incident post-mortems

NOT FOR:

• Full codebase analysis → use umlforge_reverse_engineer

• Async/event-driven messaging → use umlforge_event_driven

• Frontend component interactions → use umlforge_frontend_components

• Database schema design → use umlforge_erd_schema

Produces:

• Sequence diagram: happy path + at least 2 failure paths, activation boxes, sync vs async arrows, performance boundary annotations

• Inter-service dependency table: caller, callee, protocol, failure mode, mitigation

• (report_mode=True) Design Score: Resilience, Performance, Contract Clarity (A–F)

Args: services: All participants, e.g. "API Gateway, Auth Service, Order Service, Payment Provider, User DB". user_journey: The action to diagram in plain English, e.g. "User places an order through checkout". report_mode: True → also produce a Design Score report. Pro/Team/Enterprise only. Omit to use your saved account preference (dashboard toggle); False forces diagrams only.

Generate a deployment topology and CI/CD pipeline diagram.

USE THIS WHEN:

• You want to visualise your cloud infrastructure and how services connect

• You need to document your CI/CD pipeline from commit to production

• You are planning infrastructure, disaster recovery, or a DevOps handover

NOT FOR:

• Application-level architecture (how code is structured) → use umlforge_reverse_engineer

• Security threat modelling → use umlforge_threat_model

• How services call each other at the API level → use umlforge_api_sequence

Produces:

• Deployment diagram: nodes, artefacts, network paths, protocol/port labels, internet-facing vs internal traffic distinction

• CI/CD pipeline flow: commit → build → test → staging → production, automated gates, manual approvals, rollback paths

• Deployment environment table: infrastructure, triggers, data classification, monitoring, rollback strategy per environment

• Infrastructure risk note: SPOFs, missing redundancy, environment parity gaps

• Observability gap flags (⚠️) for services without /health endpoints

• (report_mode=True) Infrastructure Health Report: reliability score, observability score, deployment safety score (A–F)

Args: system_name: Name of the system (e.g. "UML Forge API"). cloud_provider: Cloud provider(s) (e.g. "AWS", "Railway + Vercel", "GCP"). environments: Deployment environments (e.g. "development, staging, production"). services: Services and infrastructure (e.g. "FastAPI API, Next.js, PostgreSQL, Redis"). cicd_tool: CI/CD tool (e.g. "GitHub Actions", "GitLab CI") (optional). report_mode: True → also produce an Infrastructure Health Report. Pro/Team/Enterprise only. Omit to use your saved account preference (dashboard toggle); False forces diagrams only.

Design a database schema with ERD, integrity rules, and index recommendations.

USE THIS WHEN:

• You are designing or reviewing a database schema

• You need an ERD with foreign keys, cardinality, and data types

• You want index recommendations and N+1 risk flags

NOT FOR:

• Full codebase analysis (which may include DB) → use umlforge_reverse_engineer

• API flows between services → use umlforge_api_sequence

• Event-driven data pipelines → use umlforge_event_driven

Produces:

• erDiagram: entities with typed attributes, cardinality, FK labels

• Schema narrative: one paragraph per entity — purpose, index recommendations, denormalisation decisions

• Data integrity checklist: uniqueness, FK integrity, null policies, constraints

• N+1 query risk flags

• (report_mode=True) Design Score: Normalisation, Query Performance, Data Integrity (A–F)

Args: domain_description: What the database stores (e.g. "E-commerce: users, products, orders, payments"). entities: Known entities and key attributes (e.g. "User(id, email, tier), Order(id, user_id, status, total)"). access_patterns: Most frequent read/write queries (optional — used for index recommendations). db_technology: Database technology (default: PostgreSQL). report_mode: True → also produce a Design Score report. Pro/Team/Enterprise only. Omit to use your saved account preference (dashboard toggle); False forces diagrams only.

Design an event-driven or async messaging architecture (Kafka, SQS, RabbitMQ, etc.).

USE THIS WHEN:

• Services communicate via events or messages (not direct API calls)

• You are designing event sourcing, CQRS, or pub/sub patterns

• You want to model producer → broker → consumer flows with failure handling

NOT FOR:

• Synchronous REST/gRPC calls between services → use umlforge_api_sequence

• Entity lifecycle (Order goes pending → active) → use umlforge_state_machine

• Full system architecture overview → use umlforge_stakeholder_arch

Produces:

• Event flow sequence: producers → broker → consumers with ack, retry loops (max N), dead-letter queue handling

• Event catalogue table: name, producer, consumers, payload, idempotency, retention

• Choreography vs orchestration assessment with coupling risk flags

• Failure mode analysis: scenario, impact, detection, recovery

• (report_mode=True) Event System Analysis Notes: reliability risks, idempotency gaps, resilience quick wins

Args: system_context: What this event-driven system does and why it uses messaging. producers: Services that emit events (e.g. "Order Service emits order.placed"). consumers: Services that consume events (e.g. "Notification, Inventory, Analytics"). broker: Message broker (e.g. "Kafka", "RabbitMQ", "AWS SQS/SNS") (optional). events: Named domain events (e.g. "order.placed, payment.failed") (optional). report_mode: True → also produce Event System Analysis Notes. Pro/Team/Enterprise only. Omit to use your saved account preference (dashboard toggle); False forces diagrams only.

Design a frontend component tree and interaction flow for a UI feature.

USE THIS WHEN:

• You are designing or documenting a React, Vue, Angular, or Svelte feature

• You want to see component hierarchy, props flow, and state management

• You need to identify god components or prop drilling issues

NOT FOR:

• Backend service interactions → use umlforge_api_sequence

• Full codebase including frontend → use umlforge_reverse_engineer

• Overall system architecture for stakeholders → use umlforge_stakeholder_arch

Produces:

• Component hierarchy graph: parent-child, props (downward arrows), events/callbacks (upward dashed arrows), state store connections, API origins

• Interaction sequence diagram: most complex user flow — loading, success, error

• Component responsibility table: responsibilities, state owned, reusability flag

• Accessibility note: ARIA roles needed, keyboard nav, WCAG risks

• God component flags (⚠️) for components with too many responsibilities

• (report_mode=True) Component Analysis Notes: coupling risks, refactoring quick wins

Args: feature_description: The UI feature or page to diagram. framework: Frontend framework (default: React). Supported: React, Vue, Angular, Svelte, Next.js. state_management: State approach — Redux, Zustand, Context API, Pinia (optional). interactions: Key user interactions and data flows (e.g. "user submits form, table sorts on header click") (optional). report_mode: True → also produce Component Analysis Notes. Pro/Team/Enterprise only. Omit to use your saved account preference (dashboard toggle); False forces diagrams only.

Retrieve the result of a UML Forge generation that was still running.

USE THIS WHEN:

• Another umlforge_* tool replied "still generating" and gave you a job id

• You want to fetch a result you started earlier (kept for one hour)

NOT FOR:

• Starting a new diagram → call the umlforge_* diagram tool you need

• Choosing a tool → use umlforge_suggest

Waits up to about 45 seconds for the job. If it is still running, it replies "still generating" again with the same job id: call this tool again.

Args: job_id: The job id from the "still generating" reply.

Visualise a legacy codebase and produce a migration roadmap to a modern language.

USE THIS WHEN:

• You have COBOL, Fortran, ABAP, RPG, PL/I, Pascal, BASIC, Assembly, Ada, or ALGOL code

• You want to understand the structure of a legacy program before migrating it

• You need diagrams showing what the system does NOW and what it would look like rebuilt in Python, Go, Rust, Java, TypeScript, C#, or Kotlin

• You are planning or pitching a modernisation project and need a migration roadmap

NOT FOR:

• Analysing modern codebases → use umlforge_reverse_engineer

• Generating translated source code (Option B — future feature)

• Database schema design → use umlforge_erd_schema

Produces:

• Legacy structure diagram: program divisions, sections, modules, data stores, file I/O, external interfaces (as the system exists today)

• Modern equivalent diagram: proposed clean-architecture rebuild in the target language with idiomatic layer names and structure

• Migration roadmap: 6-phase plan (Understand → Extract → Data → Logic → Test Parity → Cutover) with duration estimates and exit criteria

• Complexity & risk flags: global state, file I/O coupling, implicit typing, unstructured control flow, vendor extensions, interface contracts

• (report_mode=True) Legacy Migration Assessment: complexity scores, recommended strategy, tooling recommendations, risk assessment (A–F per dimension)

Provide EITHER github_url OR legacy_code — not both.

Args: legacy_code: Paste the legacy source code directly. github_url: Public GitHub URL containing legacy source files. Accepted formats: github.com/owner/repo, github.com/owner/repo/tree/branch/path, github.com/owner/repo/blob/branch/file.cbl source_language: Language of the legacy code (default: COBOL). Supported: COBOL, Fortran, ABAP, RPG, PL/I, Pascal, BASIC, Assembly, Ada, ALGOL. target_language: Modern language to migrate toward (default: Python). Supported: Python, Go, Rust, Java, TypeScript, C#, Kotlin. system_purpose: Brief description of what the program does in production (e.g. "monthly payroll batch", "order entry OLTP"). Optional but improves diagram labels and migration advice. github_token: GitHub personal access token (repo scope) — only needed for private repositories. Never stored or logged. report_mode: True → also produce a Legacy Migration Assessment with complexity scores and tooling recommendations. Pro/Team/Enterprise only. Omit to use your saved account preference (dashboard toggle); False forces diagrams only.

Update diagrams you already have to reflect what changed in a sprint or PR.

USE THIS WHEN:

• You ALREADY HAVE Mermaid diagrams (from a previous generation or your own)

• Code has changed in a sprint or PR and your diagrams are now out of date

• You want a changelog-annotated diff of your diagrams

NOT FOR:

• Generating diagrams for the first time → use umlforge_reverse_engineer (for existing codebases) or umlforge_stakeholder_arch (for new designs)

• Analysing a GitHub URL → use umlforge_reverse_engineer instead

• Generating diagrams without existing ones to update → use any other tool

IMPORTANT: current_diagrams is REQUIRED — paste your existing Mermaid diagrams (including the ```mermaid fences). This tool cannot generate from scratch.

Produces:

• Updated diagrams for affected sections only (with %% changelog headers)

• New diagrams for newly introduced patterns

• Architecture evolution note for pasting into a wiki

• (report_mode=True) Documentation Analysis Notes: change impact, drift risks, coverage quick wins

Args: current_diagrams: Your existing Mermaid diagram(s) — paste the full content including ```mermaid fences. REQUIRED. sprint_changes: What changed this sprint: new components, removed flows, renamed services, modified behaviour. affected_files: Files or modules touched in this sprint/PR (optional). report_mode: True → also produce Documentation Analysis Notes. Pro/Team/Enterprise only. Omit to use your saved account preference (dashboard toggle); False forces diagrams only.

Create a day-one knowledge-transfer package for a developer joining a team.

USE THIS WHEN:

• A developer is joining a new team or project and needs to get up to speed

• You are handing off a module or system to another team

• You want gotchas, constraints, and workflow diagrams in one package

NOT FOR:

• Analysing existing code for architectural problems → use umlforge_reverse_engineer

• Designing a new system → use umlforge_stakeholder_arch

• Documenting a specific API flow → use umlforge_api_sequence

Produces:

• System overview (C4 Container): the lay-of-the-land on day one

• Developer workflow sequence: local dev → test → CI → staging → production

  • most common debugging path

• Gotchas & constraints table: what the code does, why, what breaks if changed

• (report_mode=True) Onboarding Analysis Notes: coverage assessment, knowledge gaps, documentation quick wins

Args: system_description: High-level description of the system. tech_stack: Technologies in the stack (e.g. "FastAPI, PostgreSQL, React, Railway"). key_workflows: 2–3 flows a new developer must understand first. pain_points: Known gotchas, non-obvious decisions (optional). report_mode: True → also produce Onboarding Analysis Notes. Pro/Team/Enterprise only. Omit to use your saved account preference (dashboard toggle); False forces diagrams only.

Analyse an existing codebase and produce UML class, sequence, and state diagrams.

USE THIS WHEN:

• You have a GitHub URL and want to understand what the code does

• You want to audit a codebase for architectural problems

• You need diagrams of code that already exists (your own or a public repo)

• You want an Architectural Intelligence Report on any codebase

NOT FOR:

• Designing a new system from scratch → use umlforge_stakeholder_arch

• Updating diagrams after a sprint → use umlforge_living_docs

• Documenting database schema → use umlforge_erd_schema

• Mapping how services call each other → use umlforge_api_sequence

Produces:

• Class diagram: entities, attributes, relationships, multiplicities

• Sequence diagram: primary execution flow or dominant use case

• State diagram: entity lifecycle (if stateful entities are detected)

• Architectural smell flags: god classes, circular deps, anemic models

• (report_mode=True) Architectural Intelligence Report: system overview, key findings, modernisation roadmap, health scores (A–F)

Provide EITHER github_url OR codebase — not both.

Args: github_url: Public GitHub URL. Accepted formats: github.com/owner/repo github.com/owner/repo/tree/branch/path/to/dir github.com/owner/repo/blob/branch/path/to/file.py codebase: Paste code directly when you have files in context or the repo is private. max_nodes: Max classes/components per diagram (default 20, range 5–50). github_token: GitHub personal access token (repo scope) — only needed for private repositories. Never stored or logged. report_mode: True → also produce an Architectural Intelligence Report. Pro/Team/Enterprise only. Omit to use your saved account preference (dashboard toggle); False forces diagrams only.

Generate a C4 architecture diagram for a mixed technical/non-technical audience.

USE THIS WHEN:

• You are designing or presenting a system to stakeholders (CTO, investors, board)

• You need a high-level picture of a system you are planning or explaining

• You want Context + Container + Component diagrams in one call

NOT FOR:

• Analysing code that already exists → use umlforge_reverse_engineer

• Tracing a specific API request across services → use umlforge_api_sequence

• Infrastructure and deployment topology → use umlforge_deployment

• Security threat modelling → use umlforge_threat_model

Produces:

• C4 Context: system boundary, external actors, primary integrations

• C4 Container: deployable units, tech stack labels, communication protocols

• C4 Component: internals of the most complex container

• (report_mode=True) Architecture Communication Notes: audience fit, communication gaps, clarity quick wins

Args: system_description: What the system does, who uses it, its major components. audience_description: Who will read this (e.g. "CTO and non-technical board", "backend engineers new to the system"). report_mode: True → also produce Architecture Communication Notes. Pro/Team/Enterprise only. Omit to use your saved account preference (dashboard toggle); False forces diagrams only.

Design a state machine for a domain entity that has a lifecycle.

USE THIS WHEN:

• An entity moves through states (Order: pending → active → cancelled)

• You need to model a workflow, approval chain, or subscription lifecycle

• You want to find missing states, invalid transitions, or race conditions

NOT FOR:

• Flows between services (requests, responses) → use umlforge_api_sequence

• Async event messaging between services → use umlforge_event_driven

• Full codebase analysis → use umlforge_reverse_engineer

Produces:

• stateDiagram-v2: all states, entry/exit actions, guard conditions, composite states, explicit ERROR and TERMINAL states

• State transition table: current state → event → guard → next state → action

• Implementation notes: DB write requirements, domain events, race condition guards

• (report_mode=True) Analysis Notes: transition risks, unreachable states, quick wins

Args: entity: The domain entity (e.g. "Order", "Subscription", "JobApplication"). states: Known lifecycle states (e.g. "pending, active, suspended, cancelled"). events: Triggers that cause transitions (e.g. "payment_received, user_cancels"). business_rules: Constraints on transitions (optional). report_mode: True → also produce State Machine Analysis Notes. Pro/Team/Enterprise only. Omit to use your saved account preference (dashboard toggle); False forces diagrams only.

*** START HERE if you are unsure which tool to use. ***

Describe what you want to achieve in plain English. This tool will:

1. Identify the right tool for your goal

2. Return the EXACT tool call with parameters pre-filled — ready to execute

3. Explain why this tool fits and suggest an alternative

USE THIS WHEN:

• You are not sure which tool to use

• You want to describe a goal ("analyse this repo", "model our order lifecycle") and get a ready-to-run recommendation

• You want to avoid trial and error with the wrong tool

Example inputs:

• "analyse https://github.com/pallets/flask and give me an architecture report"

• "show how our Order entity moves through states"

• "diagram our Kafka event flow between Order, Inventory, and Notification services"

• "I have existing diagrams and my sprint changed the auth service"

The response will include the exact tool name, all parameters pre-filled with values derived from your description, and a one-line explanation.

Args: task_description: Plain-English description of what you want to achieve. Include any relevant details: URLs, entity names, service names, tech stack — the more context, the better the pre-filled parameters.

Generate a full STRIDE security threat model for a system.

USE THIS WHEN:

• You want to identify security vulnerabilities before launch

• You need to document auth flows, trust boundaries, or sensitive data handling

• You are preparing for a penetration test or compliance audit (GDPR, SOC2, PCI-DSS)

NOT FOR:

• General architecture review → use umlforge_reverse_engineer with report_mode=True

• Deployment and infrastructure topology → use umlforge_deployment

• API flow design (without security focus) → use umlforge_api_sequence

Produces:

• Auth flow sequence diagram: all failure paths, trust boundary annotations

• Data flow diagram: sensitivity labels (PUBLIC / INTERNAL / CONFIDENTIAL / SECRET)

• STRIDE threat table: all 6 categories with likelihood, mitigation, status

• Critical flags (🚨) for high-risk gaps

• (report_mode=True) Security Assessment Report: threat landscape, critical vulnerability deep-dives, compliance status, remediation roadmap, risk score

Args: system_description: What the system does, how users access it, main components. auth_mechanism: Auth in use (e.g. "JWT Bearer token", "API Key", "OAuth2 + PKCE"). trust_boundaries: Boundary crossings (e.g. ["public internet → API", "API → database"]) (optional). sensitive_data: Sensitive data types (e.g. ["user emails", "payment tokens"]) (optional). compliance_framework: Compliance scope (e.g. "GDPR", "NDPA 2023", "PCI-DSS") (optional). report_mode: True → also produce a Security Assessment Report. Pro/Team/Enterprise only. Omit to use your saved account preference (dashboard toggle); False forces diagrams only.