Esset is the CRM you run by talking to it. The agent creates the record types and fields your work actually needs, files your mail and meetings against them, and keeps the model growing as the job changes.
Highlights
Recent activity
View allThe agent update, annual_revenue, estimated_num_employees, primary_location
Aug 24, 2026, 11:04 AMDana Whitfield update, engagement_status, next_action_date
Aug 22, 2026, 4:37 PMThe agent create, name, domains, industry, categories, linkedin_url, website_url, technologies, annual_revenue, primary_location, founded_year
Aug 19, 2026, 9:12 AMRelated
A company record in Esset. Every attribute here is one the agent can read, write, or create.
People, companies, deals, pipelines, tasks and notes. Gmail and Calendar file themselves against the right record.
What no other CRM's agent does: it changes the shape of the database itself. New record types and fields link together as the work turns up.
Built for the size where everyone knows every deal. Not fifty reps across three regions.
Gmail and Google Calendar file correspondence and meetings against the right record. Choose OpenRouter, OpenAI, Anthropic or Google AI Studio and use your own key.
Run it on OpenRouter, OpenAI, Anthropic or Google AI Studio, on your own key.
Talk, Manual and Auto can be changed mid-conversation without opening settings.
No credit meter and no per-seat AI allowance. Nothing here is metered by us.
The agent creates the record types and fields as the work turns up.
Here it is where you think: work out who to approach, draft, revise, then send. Each person gets their own email. Hand it a file and it reads the columns, proposes the mapping, and waits for you to say yes.
I'll check the workspace schema first to see what fields exist on people and companies, then we can map this Apollo export.
Now I can see the workspace schema. Let me check if any of these companies or people already exist in the CRM before proposing imports.
The Apollo export has at least 6 contacts across 5 companies. Daniel Reyes already exists in the CRM (with email and phone). The other 5 people and all 5 companies are new.
Mapping plan (Apollo → CRM):
| Apollo field | CRM field |
|---|---|
| Company Name | company.name |
| Website | company.website_url |
| Company LinkedIn | company.linkedin_url |
| # Employees | company.estimated_num_employees |
| Industry | company.industry |
| Keywords (first few) | company.categories |
People, companies and deals ship built in, and any record type the agent creates appears here beside them with the same tools. The agent is a second way in, not the only way in.
| Name | Engagement status | Company | Email addresses | ||
|---|---|---|---|---|---|
| Amara Osei | contacted | Northwind Logistics | amara.osei@northwind.co | ||
| Bennett Cole | new | Halcyon Systems | - | ||
| Chloé Marchand | new | Vireo Analytics | c.marchand@vireo.io | ||
| Daniel Reyes | contacted | Brightpath Health | d.reyes@brightpath.health | ||
| Elena Vasquez | new | Copperline Freight | - | ||
| Farid Haddad | new | Sable & Roe | farid@sableroe.com | ||
| Grace Lindqvist | new | Torvald Marine | g.lind@torvald.se | ||
| Hiro Tanaka | new | Kitefoot Robotics | hiro@kitefoot.jp | ||
| Imani Blake | new | Ardent Foods | - | ||
| Jonas Weber | new | Meridian Textiles | j.weber@meridiantex.de | ||
| Karim Nasser | new | Lantern Group | karim@lanterngroup.ae | ||
| Lucia Ferreira | new | Verdana Estates | - | ||
| Marcus Hale | contacted | Halcyon Systems | m.hale@halcyon.io | ||
| Nadia Petrova | new | Copperline Freight | nadia@copperline.co | ||
| Oskar Lindgren | new | Torvald Marine | - | ||
| Priya Raman | new | Vireo Analytics | priya@vireo.io | ||
| Rosa Delgado | new | Ardent Foods | - |
Read, reply, compose and bulk-send from inside the CRM. The composer stays docked while you keep working the page behind it, and anything that sends to a real person still stops and asks first.
The agent inspects the schema, creates record types, adds fields with real types, and links records to other records. That is the entire difference from every other CRM agent, and everything else follows from it.
People
Companies
Deals
Sites
created from your file
The steps above are the work. This is what you get for it: a data model that grew out of the job instead of a modelling exercise you had to finish before the tool was worth anything. Every screen still exists: records, tables, boards, calendar, and inbox. The agent is a second way in.
The meeting has been queued for your approval:
Approve it and the invitation goes out from your own address.
Three things always stop and ask, on every setting including Auto:
Deleting a record is deliberately not on that list. It goes to Trash and can be restored. The list is reserved for things where undo isn't a genuine offer.
Every write is recorded in an append-only event log, so there is a record of what changed and what caused it.
People, companies and deals built in. Define your own record types and fields; they appear everywhere, including in what the agent can read and write. Records reference other records, so opening one shows everything connected to it.
Drop in a spreadsheet. Column meanings are worked out, the model is extended if needed, and you see the mapping before anything is written.
Group records into lists, read as a table or drag as a board. A record can sit on several lists holding different values on each.
Connect Gmail. Mail files itself against the person it came from. Read, reply, compose, and bulk-send.
Connect Google Calendar. Meetings appear against the record they are with. Book one from the calendar and the invitation goes out from your own address.
What's overdue and what's due today. Things you've read go quiet; things you haven't done stay loud however often you look at them.
Flagged for you, never merged behind your back.
Look up leads and companies, fill in missing detail, categorise, and generate a short brief before a call. These cost money, so they always ask first.
Attached to the record they belong to.
Deleted records are withdrawn, not destroyed. Restore them.
You bring your own API key. Pick the provider and model, and change either whenever you like. The key and the account it bills to are yours.
There is no credit meter and no ceiling. Nothing is metered by us, so you see the real cost of the work directly from your provider. That is the difference between an agent you use freely and one you ration.
No dashboards, no forecasting, no pipeline analytics. If a board asks for a monthly revenue forecast out of the CRM, this cannot produce one today. This is the biggest gap and it isn't close.
Gmail and Google Calendar. No Outlook, no Exchange, no IMAP.
No per-field access, no approval chains, no compliance exports. Everyone in a workspace sees everything.
The size where everyone knows every deal. Not fifty reps across three regions.
It is being used day to day rather than sold in volume. Some things will be rough, and some things will be missing.
If the CRM's job is to report upward, buy something else. If its job is to stop the team keeping a shadow spreadsheet, this is aimed exactly at that.
A CRM you run by talking to it. The agent extends the data model itself, so an unplanned lead list gets somewhere to live instead of a mapping exercise.
4
Model providers
3
Agent modes
0
Seat credits
0
Schemas to design
Bring your own key · Gmail & Google Calendar · Append-only event log
Neither of them is a schema. Connect what you already use, and the filing takes care of itself.
Optional. Esset is still a working CRM without it, but it cannot read your inbox.
From OpenRouter, OpenAI, Anthropic or Google AI Studio.
The agent builds the record types and fields as the work turns up.
2 thingsand no setup call