Open Agents
Sign in to the app and open the Agents area from the product switcher.
Agents docs
Amargi Agents
Amargi Agents is the AI agent runtime: define an agent, give it tools, invoke it from any Amargi product or your own systems, and observe every run with full token, cost, and tool-call traces.
Overview
Amargi Agents is a multi-tenant AI agent runtime. You define an agent with a Role, a System prompt, and a Tools list, then invoke it from anywhere: Reach AI auto-replies, Escalate AI suggestions, or your own systems via API. Every Run is captured in full detail: model invocations, input and output tokens, USD cost, latency, tool calls with their inputs and outputs, and the final decision.
We run on Anthropic Claude by default, contractually bound not to train models on customer prompts or responses. Enterprise teams can plug in their own API keys or self-host.
Quickstart
Open the app and stand up your first agent in five steps.
New agent
Click New agent. Enter the name, the role (e.g. Support Classifier), and the system prompt.
Attach tools
Pick from the built-in tool library (lookup_order, send_email, query_kb) or register custom tools via a capability manifest.
Test in the playground
Enter a message and watch the run stream live with every tool call and reasoning step.
Hook it to a trigger
Listen to inbound Reach messages, invoke from Escalate on ticket-opened, or call it from your own API.
Key concepts
Agent
A static definition that says how the agent behaves when invoked. Contains: name, role, system prompt, allowed tools list, LLM model (claude-sonnet-4, etc.), cost guardrails, and a max turn count to prevent loops.
Tool
A function the agent can invoke, defined by a JSON schema (name, description, inputs, outputs). Tools are either built-in (managed by Amargi, e.g. query_kb) or custom, where you register an HTTPS endpoint that the Agents runtime invokes when needed.
Run
A single invocation of the agent. The record contains the input, every turn of the model loop (reasoning, tool calls, tool responses), the final output, input and output tokens, USD cost, and latency. Each run writes a row to escalate.agent_runs.
Capability manifest
A registerable JSON document describing a custom tool: name, description, input schema, output schema, and an invocation URL (HTTPS, must be HMAC-signed). Lets teams add internal tools (lookup_account, charge_card, etc.) to their organization's tool library without writing code.
Approval chains
For high-stakes tools (transferring money, deleting data, sending to an external customer), set a requires-human-approval policy. When the agent tries to invoke that tool the run pauses, a task is created for a human reviewer, and the run resumes after approval or fails on rejection. Every approval is logged.
Common tasks
Create a classifier agent
- Agents, then New agent.
- Role: Support Classifier.
- System prompt: Classify the incoming support message into billing | shipping | bug | general.
- Tools: none (plain output is enough).
- Model: claude-haiku-4 (for speed and cost).
- Save, then test in the playground.
Register a custom tool
- Write a JSON manifest describing the tool (name, inputs, outputs, URL).
- Tools, then Register manifest, and paste the JSON.
- Agents generates an HMAC secret. Your tool server must verify the X-Agents-Signature header.
- Use the tool in any agent definition.
Invoke an agent from API
- Get an API key from Amargi Workspace (Settings, then API Keys).
- POST to /api/v1/agents/:id/runs with the JSON body { "input": "...user message..." }.
- The response contains run_id (for later querying) and the final output.
Monitor costs
Every run logs USD cost in the cost_usd field, aggregated automatically per day and per agent on the Agent activity card on the Reports page. Set monthly cost-threshold alerts in Settings, then AI Cost Quota.
API reference
All Agents endpoints sit under the base URL below.
https://agents.amargicreative.com/api/v1/| GET | /agents | List agents. |
| POST | /agents | Create a new agent. |
| GET | /agents/:id | Retrieve an agent. |
| POST | /agents/:id/runs | Invoke an agent (synchronous). |
| POST | /agents/:id/runs/async | Invoke an agent (asynchronous, returns run_id immediately). |
| GET | /runs/:id | Retrieve a run with full details. |
| GET | /runs | List runs. Filter by agent_id, from, to, agent_role. |
| GET | /runs/summary | Aggregate summary (runs per day, cost, tokens). |
| POST | /tools | Register a custom tool via manifest. |
| GET | /tools | List tools (built-in and custom). |
| POST | /approvals/:id/approve | Approve a paused run. |
Integration with other products
Reach
AI auto-replies invoke agents when a WhatsApp message arrives. The run shows up in the conversation log and the operator panel.
Escalate
Invokes agents on ticket.opened to suggest a resolution, shown to the operator as an AI suggestion they can accept or reject.
Your systems
Invoke from any backend server via API. Useful for ticket summarization, email classification, or any task that benefits from an LLM.
Getting help
Email support: contact@amargicreative.com. Reach out any time and we will point you to the right guide.
Privacy posture (Anthropic training, run retention): privacy posture
Explore the rest of the platform
Ready to put an agent to work?
Tell us what you want to automate and we will show you how Agents fits. Or request access and we will get you set up.