Introduction to Shara

Shara is a platform of departmental AI agents governed by a central orchestrator, Amadeus. An agent is an AI assistant that covers one area of your company (sales, finance, legal and so on). Each one works with your corporate identity, your tone and your rules, always with full auditing and human control. Instead of twenty loose chatbots, you get one coordinated team that answers to a single person.

What Shara is

Shara is a departmental AI team: a director (Amadeus) receives your requests, decides which specialists resolve them and gives you back a single, consolidated answer. Each specialist masters their area (customer service, sales, marketing, finance, HR, operations, legal, analytics and more) and acts according to the rules you define in your corporate identity.

Unlike a generic assistant, Shara is designed to delegate real operational work without losing control: every action with an effect in the real world (sending an email, issuing an invoice, placing an order) goes through your human approval before it executes, and every decision is left in a queryable audit log. The promise isn’t “a smarter chatbot”, it’s a team that works, documents what it does and asks your permission for what matters.

• A single point of contact. You talk to Amadeus the way you’d talk to an operations director; it distributes the work internally. Each employee, in turn, talks to the agent for their own area.
• On-brand by default. Everything Shara produces comes out in your tone, your values and your rules, not in a neutral factory voice.
• With a handbrake. Sensitive actions (money, contracts, external communication) require your sign-off; nothing runs behind your back.
• Traceable end to end. From the request to the result, everything is logged: who asked for what, which agent resolved it, which model was used and what was consumed.

Shara doesn’t replace your team: it takes the repetitive work off their plate and lets people decide. The human is always in the loop for the actions that matter.

Who it’s for

For SMEs without a technical team who want to delegate operational tasks to AI without giving up control. It works especially well in office sectors with a high volume of repetitive work: accounting and tax advisory firms, B2B consultancies, and marketing and communication agencies. If you need work to be logged, reviewable and under your sign-off, Shara is built for you. Below we go into detail on each of these three profiles.

The end-to-end flow

The best way to explain Shara is to follow a request from the moment it leaves your mouth to the moment you get the result. Imagine you write to Amadeus: “Prepare this week’s support ticket close-out and propose a follow-up email to the three clients with the most incidents.” Here’s what happens underneath, step by step:

1. Secure entry. Your message arrives signed with your login. Shara ties it to your company as a mandatory filter: from here on, no other client’s data is accessible.
2. Identity loading. Before reasoning, your corporate identity (IDENTITY.md, the file with your tone, values and rules) is loaded. Everything produced from here on comes out on-brand.
3. Orchestration. Amadeus analyzes the request and breaks it down: the ticket summary belongs to Wanamaker (customer service), and so does the follow-up email, with support from Patterson (sales) if there’s a commercial opportunity. It delegates with the necessary context, without leaking data between agents.
4. Model selection. Shara picks the right model by alias (Prelude, Sonata or Symphony) based on your plan, the complexity of the task, the time of day and your available quota. In the app you don’t have to choose it. Inference is served from infrastructure in the EU.
5. Execution with connectors. The agents read from your connected tools (helpdesk, CRM, email) with permissions you delegate through OAuth, the standard “sign in with” flow. Reading is direct; writing requires approval.
6. Action → approval. The follow-up email isn’t sent: it’s generated as a proposal and lands in your approval inbox. You review it, edit it if you want and approve or reject it.
7. Output review. Before returning anything, a filter reviews the response to avoid leaks (for example, the real provider behind an alias) and out-of-context sensitive data.
8. Consolidation. Amadeus merges the agents’ results into a single response with your tone, and delivers it to you along with the actions pending your sign-off.
9. Traceability and consumption. The whole chain is logged in the audit log (who, what, which agent, which model, before/after) and consumption is counted in STU, the single unit of measurement (see STU and quotas).

What you receive isn’t just the text: it’s the text + the proposed actions + the trail of how it was reached. That traceability is what turns an AI response into something you can audit, defend to a client or an inspection, and improve over time.

If you’d rather see it as data, this is a real call to the public API, POST /v1/inference. It’s the one endpoint that accepts an API key (shara_sk_live_…), and the key needs the scope for that agent. We use the alias Prelude because it works on every cloud plan: your plan sets the ceiling, and asking for a higher alias answers 422 alias_plan_mismatch. The agentSlug is the agent’s API slug (here patterson, which is Patterson); the table is in Agent catalog.

# Ask the sales agent for a task
curl -X POST https://api.sharasaas.com/v1/inference \
  -H "Authorization: Bearer $SHARA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "alias": "Prelude",
    "agentSlug": "patterson",
    "messages": [
      { "role": "user", "content": "Summarize this week’s open opportunities and draft a follow-up email to the 3 clients with the most pending proposals." }
    ]
  }'

An abridged response:

{
  "alias": "Prelude",
  "content": "12 open opportunities this week. By stage: proposal (5), negotiation (4), qualification (3). I’ve drafted the follow-up email to the 3 clients; it’s waiting for your approval.",
  "usage": { "inputTokens": 1840, "outputTokens": 612 },
  "status": "completed"
}

Notice two things. alias is always one of the four public names (never the real provider). And the email hasn’t been sent: the agent proposes it and it waits in the app’s approvals tray. An API key can’t approve it: approving only accepts the session of a person signed in to the app, so the human in the loop is a design guarantee, not an option. Read REST API and Approvals and audit.

What sets Shara apart

• Amadeus orchestrates. There are no loose agents: there’s a director who coordinates the departmental agents and answers to a single person.
• A living corporate identity. Your IDENTITY.md is loaded into every interaction and updated through patches, never through a full replacement. Every patch is logged in the audit trail.
• Public aliases, no lock-in. You work with four model aliases (Prelude, Sonata, Symphony and Concerto). When a better engine appears, we swap it underneath without your team having to relearn anything. Read No vendor lock-in.
• Predictable cost + automatic saving mode. A single unit of consumption (STU) and automatic spending limits. Outside your business hours, on the cloud plans, agents drop to a lower model to save.
• Traceability as a first-class citizen. The audit log isn’t an add-on: it’s part of the product. Every action with an effect is logged with its before/after.

Three sectors where Shara fits

Shara is horizontal (it serves any office-based SME), but there are three profiles where the fit is especially direct because their day-to-day is full of repetitive work, with its own tone and a need to leave a trail. For each one: what hurts them and how the agent team covers it.

1 · Accounting and tax advisory firms

Tax, labour, accounting and corporate law firms with a high volume of tasks that repeat every month and every quarter, and brutal peaks during campaign season (income tax, closing, quarterly filings). The bottleneck is almost never complexity: it’s volume and the customer service that surrounds it.

What usually hurts them:

• The monthly tax close and quarterly VAT eat up hours of mechanical, repetitive work.
• Customer service gets swamped during campaign season: the same emails (pending documentation, deadlines, reminders) over and over.
• Onboarding new clients is manual: requesting documents, setting up folders, registering them, explaining the process.

How Shara fits: Graham (finance) prepares reconciliations, KPIs and reporting, and leaves the issuing of any invoice or payment for your approval; Wanamaker (customer service) absorbs the flood of campaign-season emails with urgency triage and replies in your tone; Maslow (HR) and Wanamaker structure repetitive onboarding. Since every document and every email come out on-brand and everything is logged, the firm gains capacity without losing the professional control its clients expect. Recommended plan: Pro.

2 · B2B consultancies

Business, IT and management consultancies with teams of 10-50 people who sell knowledge and hours. Their product is the proposal, the analysis and the deliverable, and any minute that isn’t billable or strategic is pure cost.

What usually hurts them:

• Building commercial proposals is slow: each one is put together almost from scratch even though 70% could be reused.
• Competitive and strategic analysis is done ad hoc, with no process or living repository.
• Project management lacks visibility: a project running late is spotted too late.

How Shara fits: Patterson (sales) speeds up proposals by reusing what already worked and adapting it to the client; Rockefeller (strategy) systematizes competitive analysis (five forces, mapping, positioning) and keeps it alive instead of starting from scratch every time; Nightingale (analytics, additional agent) turns scattered data into reports with narrative; and Gantt (PMO) structures projects into milestones with early delay detection. The consultant spends their time thinking and selling, not formatting documents. Recommended plan: Max.

3 · Marketing and communication agencies

Creative, performance, social media and PR agencies running several clients at once, each with its own brand, tone and calendar. The challenge isn’t a lack of ideas: it’s producing at scale while keeping each brand distinct and reporting without eating into the margin.

What usually hurts them:

• Generating on-brand content at scale for several clients without everything “sounding the same”.
• Monthly client reporting is repetitive and eats up non-billable hours.
• Coordinating the multi-client editorial calendar without overlaps or missed posts.

How Shara fits: this is where the living corporate identity shines. Each of the agency’s clients can have its own identity, so Hopkins (marketing) and Bernays (communication) produce posts, newsletters and press releases that sound like that brand and not a generic voice; Nightingale (analytics, additional agent) automates monthly reporting with narrative; and Gantt (PMO) coordinates the multi-client editorial calendar. Human approval guarantees that nothing gets published without the account manager’s sign-off. Recommended plan: Max, if the agency adds Nightingale for reporting; Pro if it chooses its 10 agents among the selectable ones.

Sector	Main pain point	Key agents	Plan
Accounting/advisory firms	Tax/labour volume + campaign-season customer service	Graham, Wanamaker, Maslow	Pro
B2B consultancies	Slow proposals + ad hoc analysis	Patterson, Rockefeller, Nightingale, Gantt	Max
Marketing agencies	On-brand content at scale + reporting	Hopkins, Bernays, Nightingale, Gantt	Pro / Max

Shara is aimed at office sectors with repetitive work. It isn’t designed for high regulatory-risk environments like pharma. Read the agent catalog to see all 21 available.

From zero to your first conversation

1. Choose a plan and sign up. You’ll receive your activation key by email, in the format SHARA-XXXX-XXXX-XXXX-XXXX.
2. Activate the company in the desktop app. Download the app, click Activate with company key, paste the key, enter your name and a password, and follow the wizard (trade name, sector, team size, goals, tone and agents). Set your business hours: outside them, saving mode applies.
3. Invite your first employee from Team, with their role and departments. The owner and the admins can talk to Amadeus; each employee talks to their own agent. The roles are in First steps.
4. Connect your tools (email, calendar, CRM, accounting) via OAuth. Read Connectors.
5. Talk to Amadeus. For example: “Give me an executive summary of the last month: closed sales, open pipeline, pending tickets and financial alerts.” Amadeus delegates and consolidates the response.

Once inside, day-to-day work happens in the desktop app (with native notifications and the approval inbox at hand). If you also want to plug Shara into your own tools, there is the REST API. A minimal example in Python that asks an agent for a task with your API key:

import os, requests

BASE = "https://api.sharasaas.com"
headers = {"Authorization": f"Bearer {os.environ['SHARA_API_KEY']}"}

resp = requests.post(
    f"{BASE}/v1/inference",
    headers=headers,
    json={
        "alias": "Prelude",
        "agentSlug": "patterson",
        "messages": [{"role": "user", "content": "List this week’s open opportunities."}],
    },
    timeout=60,
)
resp.raise_for_status()
data = resp.json()
print(data["content"])
print("Tokens:", data["usage"]["inputTokens"], "in /", data["usage"]["outputTokens"], "out")

An API key only works for inference (/v1/inference) and for the MCP gateway. Everything else, such as your consumption (/usage/current) or the approvals inbox (/approvals), needs the session of a person signed in to the app, so a script with only an API key can’t read them.

Questions while getting started? Write to support@aiginer.com.

Where to go next

This page explains the idea in full. The start page is the short map. From here, depending on what you want to understand:

• First steps: activate your account, invite your team and leave your first task in five steps.
• Agent catalog: Amadeus + the 20 specialists, what each one does and on which plan.
• Plans and pricing: First, Pro, Max and Concerto Local, with their STU quota and agents.
• Connectors: which tools you can plug in (email, chat, calendar, CRM, accounting) and how.
• STU and quotas: how consumption is measured, what an STU is and how spending limits work.
• No vendor lock-in: why you work with model aliases and never end up tied to a provider.
• REST API: authentication, endpoints and SDKs to integrate Shara.
• Concerto Local: Shara on your own infrastructure with the Concerto model (custom pricing, with an optional 60M STU cloud bundle).
• Glossary: Shara’s terms (tenant, STU, identity, alias, orchestrator) on one page.