Gremia Builder — Creating Team Manifests¶

The Gremia Builder is a web-based chat interface where you describe your business needs in natural language. The Architect agent analyzes your input and generates a team_manifest.json that defines your entire AI operational team.

Accessing the Builder¶

Navigate to the Builder at https://app.gremia.io/builder/new or click New Team from the dashboard.

Authentication required

You must be signed in with a Gremia account. Sign up at https://app.gremia.io/register.

How It Works¶

sequenceDiagram
    participant U as You
    participant B as Builder (Web)
    participant C as Cloud (Architect)
    participant DB as Supabase

    U->>B: Describe your business
    B->>C: POST /api/v1/chat (SSE)
    C->>C: Architect graph processes
    C-->>B: Stream response chunks
    B-->>U: Display generated manifest
    U->>B: Confirm & save
    B->>C: POST /api/v1/manifests
    C->>DB: Store manifest (JSONB)

Step-by-Step Guide¶

1. Describe your team¶

Start by telling the Builder what your business does and what you need. Be as specific as possible:

Good prompt:

"I manage a property management company with 50 rental units. I need: - An agent to screen tenant applications and flag issues - An agent to handle maintenance request triage - A weekly report generator for occupancy and revenue - Integration with Google Sheets for data and Gmail for notifications"

Too vague:

"I need some AI agents for my business."

2. Review the response¶

The Architect agent responds with:

A summary of the team it will create
The agents, their roles, and the tools they will use
Suggested workflows connecting the agents
Security configuration based on your industry

3. Iterate¶

You can refine the manifest through follow-up messages:

"Add an agent for handling late payment reminders"
"Change the report schedule from weekly to daily"
"Use the restricted security level instead of standard"

Each iteration updates the manifest incrementally.

4. Preview and save¶

Once satisfied, the Builder shows a Manifest Preview with:

Summary view — Human-readable overview of agents, workflows, and tools
JSON view — Raw team_manifest.json for inspection
Download — Save the JSON file locally

Click Save to persist the manifest to the cloud. It becomes available for loading in the Gremia Shell.

Manifest Structure¶

The generated manifest follows the TeamManifest schema (validated with Zod):

Section	Description
`version`	Schema version, always `"1.0"`
`team_id`	Unique UUID for this team
`org_id`	Your organization UUID
`name`	Human-readable team name
`industry`	Industry category (healthcare, finance, retail, etc.)
`agents`	Array of agent definitions with roles, tools, and permissions
`workflows`	Directed graphs connecting agents into task pipelines
`mcp_requirements`	MCP servers needed (command, args, env, transport)
`security`	Encryption, MFA, data residency, audit retention
`ui_hints`	Dashboard layout, theme, colors
`metadata`	Timestamps, creator, manifest hash

Agent Definition¶

Each agent in the manifest includes:

{
  "id": "tenant-screener",
  "name": "Tenant Screening Agent",
  "role": "Screen tenant applications",
  "description": "Reviews rental applications...",
  "model": "balanced",
  "system_prompt": "You are a tenant screening specialist...",
  "tools": ["sheets:read_range", "sheets:write_range"],
  "permissions": {
    "allowed_tools": ["sheets:read_range"],
    "denied_tools": [],
    "max_file_size_mb": 10,
    "allowed_paths": ["/data/applications"],
    "network_access": true
  },
  "coordination": "worker",
  "approvals_required": true,
  "max_iterations": 10
}

Model Tiers¶

Tier	Speed	Cost	Use Case
`fast`	Fastest	Lowest	Simple lookups, formatting
`balanced`	Medium	Medium	Analysis, writing, decisions
`powerful`	Slowest	Highest	Complex reasoning, code generation

Coordination Roles¶

Role	Behavior
`supervisor`	Delegates tasks, reviews results
`coordinator`	Routes work between agents
`worker`	Executes assigned tasks
`validator`	Reviews and approves outputs

Supported Industries¶

The Builder tailors security policies and tool suggestions based on your industry:

Healthcare — HIPAA-compatible, restricted security, audit everything
Finance — SOC 2 aligned, encryption mandatory, approval gates
Legal — Document management, confidentiality controls
Logistics — Fleet tracking, route optimization integrations
Retail — Inventory, customer service, sales analytics
Manufacturing — Quality control, supply chain monitoring
Education — Student data protection, curriculum tools
Hospitality — Booking management, guest services
Real Estate — Property management, tenant screening
Custom — Define your own requirements

Tips for Better Manifests¶

Be specific about tools — Mention the services you use (Google Sheets, Slack, GitHub)
Define approval gates — Specify which actions need human review
Set security expectations — Mention compliance requirements (GDPR, HIPAA)
Describe workflows — Explain how agents should hand off work to each other
Iterate — Start simple and add complexity through follow-up messages