Skip to content

Understanding the FCC Ecosystem

This tutorial introduces the FCC ecosystem -- a family of 11 projects that work together to support AI-assisted software development workflows. It is aimed at beginners who want to understand how the projects relate to each other and where to start.

Table of Contents

  1. What Is the FCC Ecosystem?
  2. Project Roles
  3. Tier System
  4. How Projects Communicate
  5. Getting Started by Role
  6. Cross-Project Workflows
  7. Key Concepts
  8. Frequently Asked Questions
  9. Related Documentation

What Is the FCC Ecosystem?

The FCC (Find, Create, Critique) Ecosystem is a collection of 11 interconnected projects that implement a structured framework for AI-assisted software development. At the center is the FCC framework itself, which defines 102 personas, 6 workflow action types, and a protocol bridge that connects all the other projects.

Each project has a defined role, a tier level indicating its relationship to the hub, and a set of capabilities it exposes to the rest of the ecosystem.

The ecosystem is designed so that projects can be adopted incrementally. You do not need all 11 projects to get value -- start with the hub and add projects as your needs grow.

Project Roles

Project Role Tier Key Capability
FCC Framework hub 0 102 personas, protocol bridge, event bus
PAOM Pattern engine 1 13 personas, 25 MCP tools
CTO Object model 1 1520 concepts, full A2A support
AOME Vocabulary alignment 2 Taxonomy alignment across projects
CONSTEL Knowledge graph 2 TMF federation, graph queries
Research Center Statistics 2 8 A2A skills, statistical analysis
Sky-Parlour Visualization 2 D3.js components, A2A handlers
AI-COE-DOCS Documentation 2 MCP server for doc generation
Distiller Data fabrication 3 NanoCube engine, vocabulary mapping
FlowX Event capture 3 OpenTelemetry integration
L2 Distro Hub Distribution 3 Release management, packaging

What each project does

FCC (Tier 0) is the hub. It defines the persona specifications (R.I.S.C.E.A.R.), the workflow graphs (Find, Create, Critique), the EventBus for inter-component messaging, and the protocol bridges (A2A, MCP) that let other projects communicate.

PAOM (Tier 1) is the pattern and antipattern observation engine. It provides 13 specialized personas focused on pattern recognition and 25 MCP tools for automated pattern analysis. PAOM consumes the FCC framework -- it does not define it.

CTO (Tier 1) provides a rich object model with 1520 concepts. It offers full A2A agent card support and serves as the canonical vocabulary for domain concepts across the ecosystem.

AOME (Tier 2) handles vocabulary alignment. When two projects use different terms for the same concept (e.g., "persona" vs. "agent"), AOME provides the mapping.

CONSTEL (Tier 2) is the knowledge graph project. It uses TMF (Technology Management Framework) federation to connect knowledge across projects and supports graph queries for cross-project discovery.

Research Center (Tier 2) provides statistical analysis capabilities exposed as 8 A2A skills. Use it for quantitative evaluation of workflow outcomes, persona effectiveness, and quality gate results.

Sky-Parlour (Tier 2) is the visualization project. It provides D3.js components and A2A handlers for rendering ecosystem data in dashboards.

AI-COE-DOCS (Tier 2) is the documentation project. It runs an MCP server that generates and validates documentation artifacts.

Distiller (Tier 3) handles synthetic data fabrication. Its NanoCube engine generates hierarchical test data, and its vocabulary mapping bridges domain-specific terminology.

FlowX (Tier 3) captures and exports events using OpenTelemetry integration. It connects to the EventBus and forwards traces and metrics to observability backends.

L2 Distro Hub (Tier 3) manages releases and distribution of all ecosystem packages.

Tier System

The tier system classifies projects by their relationship to the FCC hub:

        Tier 0 (Hub)
           FCC
          /   \
     Tier 1 (Core)
     PAOM    CTO
      |       |
     Tier 2 (Extension)
     AOME    CONSTEL    Research Center
     Sky-Parlour    AI-COE-DOCS
      |
     Tier 3 (Support)
     Distiller    FlowX    L2 Distro Hub

Tier 0 -- Hub

The single source of truth for the FCC framework. All other projects depend on Tier 0 definitions (personas, workflows, events, protocols).

Tier 1 -- Core

Projects that provide foundational capabilities consumed by multiple Tier 2 projects. Tier 1 projects have deep integration with the hub but are independently deployable.

Tier 2 -- Extension

Projects that extend the ecosystem with specialized capabilities. They depend on Tier 0 and may depend on Tier 1, but they do not depend on each other.

Tier 3 -- Support

Infrastructure projects that provide cross-cutting services (data generation, event capture, packaging). They can be replaced or omitted without affecting core functionality.

How Projects Communicate

Projects in the FCC ecosystem communicate through three mechanisms:

1. A2A Agent Cards

Each project can expose its capabilities as A2A Agent Cards -- self-describing manifests that advertise skills, endpoints, and metadata. The FCC hub generates Agent Cards for all 102 personas. Other projects generate their own cards for their specialized agents.

Agent Cards enable discovery: a project can query the ecosystem for agents with specific skills.

Project A                          Project B
  |                                   |
  |-- query: "find agents with       |
  |          skill=code_review" ---->|
  |                                   |
  |<-- response: [AgentCard, ...] ---|

2. MCP Tools

The Model Context Protocol (MCP) layer exposes project capabilities as callable tools with JSON Schema input specifications. FCC provides 14 default tools; PAOM adds 25 more.

MCP tools enable invocation: one project can call another project's capabilities through a standardized interface.

Client                    MCP Server
  |                          |
  |-- call: fcc_simulate  -->|
  |   {scenario: "GEN-001"} |
  |                          |
  |<-- result: {status: ok} -|

3. EventBus Federation

The FCC EventBus provides in-process pub/sub messaging with 81 event types. Events can be forwarded across projects via the WebSocket bridge, SSE bridge, or the AsyncEventBusAdapter.

EventBus federation enables observation: projects can subscribe to events from other projects without direct coupling.

EventBus (FCC)              WebSocket Bridge            Browser / Project B
    |                            |                           |
    |-- publish(event) --------->|                           |
    |                            |-- broadcast(json) ------->|

Getting Started by Role

Software Developer

Start with the FCC hub project. Install it, run the tests, and explore the 102 personas. Use the CLI to scaffold a project and run a mock simulation.

cd l2_fcc_agent_team_ext
make venv && source .venv/bin/activate
make install-dev
make test
fcc init my-project
fcc simulate --mock

Architect

Start with the FCC hub and CTO object model. The object model provides a rich vocabulary for domain concepts, and the protocol bridge shows how projects communicate. Review the Ecosystem Governance Patterns document for federation topology and JV partnership patterns.

Data Scientist

Start with the FCC hub and Research Center. The Research Center provides statistical analysis skills, and the FCC simulation engine can generate data for analysis. Add Distiller for synthetic data generation.

DevOps Engineer

Start with FCC and FlowX. FlowX connects to the EventBus and exports traces to your observability stack via OpenTelemetry.

Documentation Author

Start with FCC and AI-COE-DOCS. Use the FCC docs-as-code generator (fcc generate-docs) and the AI-COE-DOCS MCP server for documentation workflows.

Cross-Project Workflows

Example 1: Pattern Discovery

  1. A developer asks PAOM to analyze a codebase for patterns.
  2. PAOM uses its 13 personas and 25 MCP tools to identify patterns.
  3. Results are published as events on the EventBus.
  4. FCC personas (RC, BC, SQC) review the findings in a Find-Create-Critique cycle.
  5. Sky-Parlour visualizes the results in a D3 force graph.

Example 2: Knowledge Federation

  1. A researcher queries CONSTEL for concepts related to "microservices".
  2. CONSTEL federates the query across CTO (1520 concepts) and AOME (vocabulary mappings).
  3. Results are returned as A2A skill responses with provenance metadata.
  4. The FCC EventBus logs the query and results for auditing.

Example 3: Quality Gate Evaluation

  1. A team submits a deliverable for quality review.
  2. FCC runs the deliverable through quality gates (25 configured gates).
  3. Research Center provides statistical significance testing on the scores.
  4. The results are published to the EventBus.
  5. FlowX captures the traces and forwards them to an observability backend.
  6. Sky-Parlour renders a quality dashboard showing pass/fail status.

Example 4: Synthetic Data for Testing

  1. A test engineer needs realistic test data.
  2. Distiller's NanoCube engine generates hierarchical test data via the FabricationBridgeAdapter.
  3. The data is validated against exchange contracts (version 2.1.0).
  4. AOME provides vocabulary mappings to ensure term consistency.
  5. FCC personas review the synthetic data for realism.

Key Concepts

R.I.S.C.E.A.R.

The 10-component persona specification used throughout the ecosystem:

Component Description
Role What the persona does
Input What the persona expects to receive
Style How the persona communicates
Constraints Rules and limitations
Expected Output Deliverable format
Archetype Behavioral pattern
Responsibilities Specific duties
Role Skills Technical skills
Role Collaborators Who the persona works with
Role Adoption Checklist Steps to adopt the persona

Find, Create, Critique (FCC)

The three-phase workflow cycle:

  • Find: Research and gather information. Identify what exists, what is missing, and what needs to change.
  • Create: Produce artifacts -- code, documents, designs, tests. Apply persona expertise to generate deliverables.
  • Critique: Review and evaluate artifacts against quality gates. Provide structured feedback and improvement recommendations.

Persona Categories

The 102 personas are organized into 20 categories:

Category Count Focus
core 5 Base FCC workflow roles
integration 8 Cross-system integration
governance 4 Quality and compliance
stakeholder 5 Business and user perspectives
champion 4 Team coordination
ux_visualization 3 User experience
protocol_engineering 3 Protocol design
jv_governance 3 Joint venture management
open_science 3 Open science compliance
responsible_ai 3 AI ethics
jv_collaboration 3 JV collaboration
docs_as_code 3 Documentation
privacy 3 Privacy engineering
knowledge_graph 3 Knowledge management
local_first_ai 3 Edge/local AI
data_engineering 6 Data pipelines
ml_lifecycle 9 ML operations
ml_models 11 Model development
devops 3 DevOps practices
app_development 2 Application building

Frequently Asked Questions

Do I need all 11 projects?

No. Start with the FCC hub (Tier 0). Add other projects as your needs grow. Each project is independently installable.

Can I use FCC without AI?

Yes. The simulation engine has a deterministic mock mode that requires no AI provider. All personas, workflows, and quality gates work without AI.

How do I add a new project to the ecosystem?

  1. Define your project's A2A Agent Cards.
  2. Register your project's MCP tools (if any).
  3. Connect to the EventBus via the WebSocket or SSE bridge.
  4. Register with the JV Partnership Registry for governance tracking.

What is the difference between FCC and PAOM?

FCC owns and defines the framework (personas, workflows, R.I.S.C.E.A.R.). PAOM is a consumer that adds pattern-specific personas and tools. PAOM depends on FCC; FCC does not depend on PAOM.

Where are the Streamlit apps?

The apps/streamlit/ directory contains 10 Streamlit applications for visual exploration. Run them with streamlit run apps/streamlit/<app>.py.