Molly AI & Agents

Ask. Confirm. Done.

Ask Molly and she runs the real query, answering with a live chart. Ask her to act and she creates and updates records and drafts replies — after you press Confirm. AI agents take it further: their own instructions, knowledge, tools and model, woken by events, working inside a monthly budget with an approval queue for anything that needs you.

Get started free Book a demo

chart types Molly can render: 57
actions Molly can take: 165
model providers: 4
of writes behind a confirm step: 100%

Ask

Questions in, artifacts out

Press ⌘K on any page and ask which renewals are overdue, grouped by owner. Molly runs a real query against your data and answers with a live chart — switchable between renderers, downloadable, openable in the app. Tables, files and images are artifacts too.

Schema-aware: she knows your objects and fields
Multiple artifacts per answer
Never silent — you always get a response

Act

The same hands as your team

Molly acts through the platform's action registry — the exact operations the UI and API use, gated by the same permissions. Every write shows a confirmation card first; nothing changes behind your back.

165 actions across 19 feature areas
Respects your role and field-level permissions
Confirm each write, with defer and resume

AI agents

Teammates you configure, not hire

Each agent gets instructions, a model, its own slice of the knowledge base and a tool belt you choose. It wakes when something happens, triages what it sees — acting on some, ignoring the rest — and queues anything that needs a person as a pending approval.

Instructions, model, knowledge and tools per agent
Woken by events; triage acted on or ignored
Pending approvals, waiting for a decision

Oversight

A budget, a queue and a paper trail

Every agent shows what it costs — this month against its budget, with the tokens behind it, by day and by thread — alongside its open threads, pending approvals and how often it woke. Switch one off from the list; everything it did stays in its threads.

A monthly budget per agent, with token counts
Cost by day and by thread
On or off from the agents list

Also in the box

Model picker

OpenAI, Anthropic Claude, Google Gemini or anything on OpenRouter — per organisation, switchable any time.

⌘K everywhere

Ask Molly from any page to find data, build a chart or create something.

Ask agent

Start an agent on a specific record straight from its header.

Secrets Vault

Encrypted, write-only values for agent sandboxes — agents see the names, never the values, and outputs are redacted.

Voice in, voice out

Voice notes are transcribed; replies can be spoken aloud.

Sandboxed code

Molly runs code in a sandbox and hands back charts and files as artifacts.

Works better together

Put your agents to work.

Bring your team, connect your channels and let klerc take the follow-ups — from first hello to paid invoice.