Drive your growth stack from your editor

Tell your agent to track revenue, launch a test, or ship a flag. It does the work, then drops the code into the right file in your project.

Try it in Claude Code See the platform

MCP and Agents in SessionSight expose every platform feature to coding agents (Claude Code, Cursor, Codex), so you can drive workflows, ship flags, and track revenue from inside your editor.

Just say the word.

You ask in plain language. The agent works through it for a moment, then tells you what it did. You don't read code unless you want to.

claude code
> /sessionsight track revenue when checkout completes
⏺Setting it up.
⎿ Created the Revenue goal
⎿ Installed the SessionSight goals package
⎿ Updated your checkout success handler
⏺Done. I created a Revenue goal targeting $10,000 per month, then added the tracking line to your checkout success handler so every order gets counted automatically. You'll see revenue land on your goals dashboard with your next sale.
✻ Cooked for 3m 43s

Ask, or act. Same agent. Zero learning curve.

Your sessions are screaming at you.

Users rarely tell you what's wrong. They just leave. But every session, replay, and rage-click is a signal, and there are thousands of them every day. No human could read them all. The agent does.

claude code
> /sessionsight why are signups dropping on Safari this week?
⏺Looking into this.
⎿ Pulled the signup funnel, 1,847 attempts this week vs 1,723 last
⎿ Broke it down by browser, then by device
⎿ Analyzed all 239 abandoned Safari sessions
⎿ Cross-referenced rage-click and hesitation timing across the group
⏺Safari signups dropped 23% this week. 87% of the 239 abandoned sessions stalled on the email field. Safari's autofill is covering the password field below. Want me to ship a fix behind a feature flag?
✻ Churned for 2m 18s

One prompt. Three things off your list.

The A/B test you've been putting off. The goal you've been meaning to track. The flag you keep almost shipping. Tell the agent. The change is in your code, the test is running, the goal is tracking. You didn't do any of it.

Shipped

This week
M
Created a goal Claude Code · you
just now
M
Started a split test Cursor · you
3m ago
M
Shipped a feature flag Claude Code · you
12m ago
M
Created a power-user segment Codex · you
37m ago
D
Launched a campaign Dashboard · you
4 days ago

173+ actions across 32 areas of the platform

Your agent can see anything in your account that makes sense for an agent to see. It can create and change a careful list of things. It can never delete anything, full stop.

112 things it can see
61 things it can change
0 deletes, ever
Campaigns
9 read 8 write
Workflows
9 read read only
Split tests
4 read 5 write
Feature flags
4 read 5 write
Goals
4 read 2 write
Heatmaps
5 read
Funnels
3 read
Funnel views
4 read 2 write
Sessions
6 read 1 write
Segments
3 read 3 write
Segment analytics
2 read
Visitor profiles
3 read
Form analytics
3 read 1 write
Feedback
6 read 2 write
Strategy
6 read 7 write
Properties
4 read 5 write
Teams + invites
5 read 6 write
Companies
2 read 3 write
Marketplace
5 read 1 write
Connections
2 read
Notifications
2 read 2 write
Live
1 read
Dashboard
2 read
Audit logs
1 read
Plans + usage
2 read
Insights
1 read 1 write
Secrets + keys
3 read 3 write
Onboarding
2 read 2 write
Billing
2 read
Me
1 read 1 write
Notification prefs
1 read 1 write
Purchased workflows
5 read

Two commands in Claude Code. The browser handles the rest.

Install

Distributed straight from the SessionSight SDKs repo. Add the marketplace, install the plugin, then ask for what you want. We handle the rest.

1
In Claude Code, run /plugin marketplace add SessionSight/sdks /plugin install sessionsight@sessionsight-sdks
2
Ask for something /sessionsight what are my goals? A browser tab opens once so you can pick which company and properties the agent can touch. Every prompt after that runs instantly.
Also works in Cursor Codex Claude Desktop Any MCP client

Safety guarantees

Coding agents are powerful. Mis-prompted ones are too. SessionSight's MCP is built so a bad prompt cannot become a bad outcome.

✓
No deleting, ever Even if your login leaked, no one could use the agent to delete a goal, archive a campaign, or remove a teammate.
✓
Per-property access When you first connect the agent, you pick which of your properties it can touch. A test setup pointed at staging can't reach production. Other tools only let you choose at the account level. We let you choose property by property.
✓
Revoke from one screen Every connected agent shows up in your account with the last time it was used. One click cuts it off. No grace period.

Stop alt-tabbing to do your marketing.

Install the plugin, ask for what you want, and watch your marketing come to life.

Try it in Claude Code
Keep exploring

Three more tools worth a look.

Session Replays

Watch the exact moment things went wrong.

Heatmaps

See where attention actually goes.

Funnels

Find the step where you're losing people.