Getting Started

CRM Agent is a CRM on your server in which an AI agent keeps the data. The agent is a full participant in working with customers: it creates and cleans the client base, keeps the timeline of communication, merges repeats, accepts requests from the site, moves deals through the pipeline, prepares quotes and invoices, assembles mailings and watches request deadlines. You watch the work and give commands: you show what needs to be done and check the result.

What CRM Agent is

It is not a program where you press buttons and fill in forms. The idea is different: the agent keeps the data, while the human watches and commands. The interface does not have a single editing form — that is not something unfinished, it is protection and the product idea. An accidental edit simply has nowhere to happen: every record is shown read-only, and next to it sits a button that inserts a ready phrase into the chat with the agent or copies a management command (the program address your agent uses).

This is deliberate. Keeping a client base is hundreds of small things (what the client’s name is, which phone is the main one, which company stands behind it, what stage the deal is at), and in forms you would be clicking through them for hours. Instead you speak in words: “create these clients from the old spreadsheet”, “merge the duplicates”, “prepare an invoice for this deal” — and the agent deals with the details. And every step it takes leaves a trace: it is visible who changed which field, when and why.

Let’s be honest right away: without a connected AI agent the data does not move by itself. Until the agent is connected, you can watch and export, but not run the work. Details in the Chat with the AI Agent section.

Who it fits

  • A sales department — instead of typing in requests and notes by hand: the agent creates contacts and companies itself, moves deals through the pipeline and reminds about the next step. See Leads and Deals & Pipeline.
  • Service and support — requests from all channels gather in one place, reaction deadlines are under control, the client’s history is in front of your eyes. See Service: Tickets, SLA, Knowledge & Portal.
  • Marketing — mailings, audiences, consents and honest statistics: it is visible which channel the clients and the money came from. See Marketing.
  • A small business without a CRM administrator — nobody needs to be taught to fill in fields: you explain the task in words, and the agent takes the routine on itself.

How the work is arranged

The client’s path from first touch to payment looks like this — each step has its own documentation page:

  • The client base — contacts and companies with their details and the links between people and organizations: Contacts & Companies.
  • The timeline — the whole history of communication with the client in one feed: notes, calls, emails, tasks, files: Timeline & Links.
  • Data quality — mandatory duplicate search, the review queue, merging duplicates and bringing the data to one form: Data Quality & Duplicates.
  • Leads — intake of requests from all channels, assignment, scoring and conversion into a client: Leads.
  • Deals — the pipeline, stages, the transition log, forecast and analytics: Deals & Pipeline.
  • Money — quotes, invoices, payments, contracts and subscriptions: Money: Quotes, Invoices, Payments & Contracts.
  • Marketing — campaigns, audiences, mailings and attribution: Marketing.
  • Communications — the shared inbox, email, calls, messengers and the site widget: Communications.
  • Service — tickets, deadlines, the knowledge base and the client portal: Service: Tickets, SLA, Knowledge & Portal.

On every step the data is entered and maintained by the agent — not through the screens but through management over the API (the product’s program addresses). That is exactly why the interface stays light: it shows instead of asking you to fill things in. How this works technically is described in the Management API section.

What the product consists of

The product is assembled from a small core and add-ons. The core holds the client base, the timeline, data quality, import and export, custom fields, management and the screens. The remaining contours — Deals, Money, Leads, Marketing and the communications and service add-on families — are switched on by choice: you pay attention only to what you really need. A fresh installation has only the core available; the rest is offered in the Addons catalogue.

What you need before the start

  • A server or a computer with Docker — Windows, Linux or macOS all work. Details in the Installation section.
  • Your Xedant Agent with an AI model — it keeps the data by your commands and serves as the chat right inside the interface. Without it everything works except the chat and the automatic maintaining of data.
  • The management key — with it the agent receives the right to change the data. It is set once at installation. Until the key is set, changes are refused, while reading keeps working.
  • Data to carry over, if you already have some — an export from an old CRM or a plain spreadsheet. The agent will carry it over and clean it. See Import, Export & Bulk Operations.

First sign-in

The first user to register becomes the administrator — there is no “default” account. If you prefer to set the sign-in in advance, give the login and password in environment variables at installation: then the first sign-in happens with those. A session lives 90 days, so you will not have to re-enter the password. If you have an external sign-in configured (single sign-on through a proxy server), users are created automatically on first visit.

First steps

  1. Connect the agent — give the address and key of your Xedant Agent. After that the chat appears in the interface, and it is where you will give your commands.
  2. Check the management key — it is set at installation and opens the right to change the data for the agent. If it is not set, changes will not go through: the product will say so honestly.
  3. Ask it to create the clients — “carry the clients over from this spreadsheet and remove the repeats”, or, if there is no base yet, “create these clients”. The agent will walk the rows, find the duplicates and show what came out.
  4. Set up the deal pipeline — “make the stages such and such”. The pipeline and the stages are data, not a hard-wired process, so they can be adjusted to yourself at any moment.
  5. Switch on the add-ons you need — “switch on Leads, Deals and Money”. Switching on applies on restart — that is more reliable than changing the composition on the fly. See Add-ons & Product Composition.
  6. Start accepting requests — take the public intake address from the Leads section and put it into a form on your site. Requests will land in the base and be assigned by themselves. See Leads.

What the product does not do

Honestly and right away, so there are no disappointments:

  • There are no editing forms in the interface — not one. The single exception is the administrative management of add-ons (install, disable, remove).
  • Without the agent the data does not move — a connected Xedant Agent with an AI model is required for the work, not for looks.
  • Part of the contours are independent add-ons — they need to be switched on and the product restarted after that.
  • There is no autonomous selling without a human — the product helps you sell, but the decisions and the closing of deals stay with the human.

Next → Contacts & Companies

← Documentation table of contents