Question 1

What is Recon, exactly?

Accepted Answer

A verified record of how your company works. Recon reads your internal docs, conversations, tickets, help desk, code, and database, resolves what it finds into beliefs with a receipt on every line, and surfaces the places your sources disagree instead of guessing. Your team and your AI tools read the same record.

Question 2

Where does Recon get its information?

Accepted Answer

From the systems you already run: internal docs, meeting transcripts, Slack threads, tickets, help desk conversations, your codebase, your database, and the rest of your stack. Connections are read-only by default and scoped to one workspace.

Question 3

What does "verified" actually mean?

Accepted Answer

Every belief Recon holds is linked to the exact source line it came from, so any answer can be traced back and checked. There is no black box to take on trust. When the source changes, the belief updates and the receipt points to the new line.

Question 4

How does it handle contradictions?

Accepted Answer

It does not average them away. When two sources disagree, the belief is marked disputed and shown with the receipts on each side. A person resolves it, and the resolution writes back as the current truth, with the earlier versions kept in history.

Question 5

Who on my team can see what?

Accepted Answer

Access is scoped to the workspace and enforced with row-level security, so people only read their workspace's record. Sources you bring in privately stay private to you; shared sources are readable by the workspace. Every read and write is captured in an audit log.

Question 6

Can my AI tools actually use it?

Accepted Answer

Yes. Recon exposes a standard read-only MCP server, so Claude, ChatGPT, Cursor, Codex, and any MCP-compatible client read the same record with the same receipts. Reads through MCP are free.

Question 7

Can Recon change my systems?

Accepted Answer

Not on its own. Integrations are read-only by default, enforced at the database role, the tool layer, and the system prompt. Write actions, like creating a Linear issue or drafting a message, are held in an approval queue and only run when you approve them. Nothing is auto-executed.

Question 8

What are Agents?

Accepted Answer

Agents are recurring tasks you point at the accounts or topics you care about. On a schedule, an Agent reviews what changed, updates the record, and proposes actions with evidence attached in your Inbox, which you approve or reject. They read the same verified record everything else does.

Question 9

What about skills, and agents that act on their own?

Accepted Answer

That is the direction we are building. The verified record (memory) is shipped today. Skills capture how your people actually work, the repeatable plays a team runs again and again. On top of that, agents act on what changes in the record and draft the work for you, still holding it for your approval. Memory first, then skills, then agents.

Question 10

Isn't this just Claude Code with my own MCP servers connected?

Accepted Answer

Claude Code is great for the engineer who set it up. Recon is for everyone else on the team. Your support or ops teammate does not write SQL, configure MCP servers, or get direct production access. Recon connects your sources once, keeps the verified record current, controls who can read what, and lets the whole team and their AI tools work from the same cited answers.

Question 11

How is this different from a memory layer like Mem0?

Accepted Answer

Mem0 is memory infrastructure for developers: you pass it conversation history through an SDK and it stores what you give it. Recon reads from the systems you already run, resolves what it finds into beliefs with a receipt on every line, surfaces contradictions, and controls who can read what. The verified record is the product, not an SDK you wire up.

Question 12

Do I need my own AI provider key?

Accepted Answer

No. Recon runs on its own managed AI provider keys, so there is nothing for you to bring or configure. Your data is never used to train any model.

Question 13

How does pricing work?

Accepted Answer

Reads through the MCP server and the initial setup are free. Heavier work draws from a monthly allowance rather than a per-seat bill. We are onboarding early teams directly and finalizing the plan structure with them, so there is nothing to commit to today. Book a call and we will set it up with you.

Question 14

How does Recon handle security?

Accepted Answer

Connection credentials are encrypted at rest with AES-256-GCM. Work runs in isolated, ephemeral sandboxes. Access is workspace-scoped with row-level security, reads are read-only by default, writes need approval, and everything is captured in an append-only audit log. Your data is never used for model training.

Question 15

What integrations do you support?

Accepted Answer

PostgreSQL, GitHub, Linear, Jira, Confluence, Notion, Intercom, HubSpot, Sentry, Zendesk, HappyFox, Stripe, and Supabase. Slack as a delivery channel so your team can ask questions from any thread. Most integrations connect via OAuth in a few clicks.

Question 16

Does my engineering team need to set this up?

Accepted Answer

Just for the initial connection. An engineer connects your sources through the dashboard once, with read-only access where applicable. After that, the rest of the team reads the shared record themselves, without routing every question back through engineering.

Question 17

We're a small team. Is this for us?

Accepted Answer

Yes. Recon is built for small B2B SaaS teams where the answers to how the company works still live in a few people's heads and a dozen tools. It puts that on a shared, cited record before the team outgrows what one or two people can hold.

Question 18

Where can I find setup instructions and documentation?

Accepted Answer

The help center covers everything: getting started, connecting each data source, installing Slack, and example questions you can ask.

Common questions.

Give your coding agent a work memory.