UserProofPostHog MCP vs UserProof: what each hands your coding agent
Disclosure: UserProof publishes this page and sells one of the two tools compared. Every PostHog fact below links to PostHog’s own docs, read October 26, 2026. We did not hands-on test PostHog’s MCP server for this page.
PostHog’s MCP server gives a coding agent the whole PostHog platform: 8 session replay tools, 52 Replay vision tools, 100 Signals tools and dozens of other categories, free to call. UserProof’s MCP server has 13 tools built around one job: hand the agent the single fix that most moves your goal, with the file and line, then check whether it worked. Pick PostHog’s when your agent should explore your data. Pick UserProof’s when it should be told what to change.
The count side by side: PostHog’s tools reference lists 8 tools for session replays, 52 for Replay vision and 100 for Signals, across more than 70 categories. UserProof ships 13 tools in total, and replay at 50,000 a month costs $29 on UserProof Pro against $172.50 on PostHog by our arithmetic from its published tiers.
Jump to: job by job · PostHog’s replay tools · UserProof’s tools · setup · cost · which to pick · FAQ
Job by job
What a coding agent can do with each server, matched to the job. PostHog cells come from its MCP docs and tools reference; UserProof cells from our MCP server source.
| Job | PostHog MCP | UserProof MCP |
|---|---|---|
| Find sessions with a problem | query-session-recordings-list: filter by events, person properties, rage and dead clicks, time range | get_top_fix returns the finding with the sessions that prove it |
| Read one session | session-recording-get: duration, interaction counts, console log counts, person summary | Sessions come attached to a finding; there is no raw session browser |
| AI review of recordings | Replay vision: 52 tools for AI scanners, observations and backfills; LLM tools bill as PostHog AI spend | No LLM in the counts; your own agent writes the explanation |
| Decide what to change | Signals: 100 tools for Inbox reports and scouts; Inbox opens pull requests, first 3 a month free then $15 each (pricing page) | get_top_fix: one fix ranked by your goal, with the file and line and dollar math when a conversion event is set |
| Set the goal | Marketing analytics conversion goals; Experiments has 50 tools | set_goal |
| Check whether it worked | Run an experiment and read its metrics | mark_fix_applied, then check_fix compares the goal before and after |
| Before you have traffic | Not covered on the pages we read | scan_site: free scan of the live site for broken elements |
| Pages missing the snippet | Not covered on the pages we read | check_coverage crawls your sitemap |
| Many sites at once | Organization and project switching tools | fleet, list_sites, select_site |
| Everything else | Feature flags (29), error tracking (28), logs (22), surveys (11), SQL and more | Nothing else: recording, findings and fix checks only |
PostHog’s 8 session replay tools
From the tools reference, read October 26, 2026: query-session-recordings-list, session-recording-get, session-recording-delete, session-recording-bulk-delete, session-recording-playlist-create, session-recording-playlist-get, session-recording-playlist-update and session-recording-playlists-list. The replay MCP page says deletion is permanent and capped at 100 recordings per batch, and that “your agent should confirm with you first.” It lists the metadata a recording read returns and does not say whether the agent sees the replay’s events.
UserProof’s 13 tools
get_top_fix, check_fix, mark_fix_applied, set_goal, scan_site, check_coverage, wait_for_first_event, get_activation, get_usage, fleet, list_sites, select_site and upgrade. When no finding passes the evidence bar, get_top_fix says so and returns the biggest visible leak in the funnel as a lead, labeled as unproven.
Setup
| PostHog MCP | UserProof | |
|---|---|---|
| Command | npx @posthog/wizard mcp add, or the hosted URL https://mcp.posthog.com/mcp | Claude Code: claude plugin install userproof@userproof after adding the marketplace; Codex: codex mcp add userproof -- npx -y github:GetUserProof/userproof |
| Clients named | PostHog Desktop, Cursor, Claude Code, Claude Desktop, Codex, VS Code, Zed, Windsurf; Lovable, Replit, v0 | Claude Code, Codex, Cursor |
| Data needed first | A PostHog project with replay turned on | The UserProof snippet; the setup skill adds it to your code |
| Calling tools | Free; LLM-backed tools need AI data processing on and bill as AI spend | Free on every plan; plans are priced on recorded visits |
Cost of the replay data behind it
| Monthly volume | PostHog replay | UserProof |
|---|---|---|
| 5,000 | $0 (first 5,000 free) | $0 on Free |
| 50,000 | $172.50 (our arithmetic from the tiers) | $29 on Pro |
| 250,000 | $542.50 (our arithmetic) | $99 on Scale |
Full tier math and retention are on our PostHog alternative page.
Which to pick
| If | Pick |
|---|---|
| You already run PostHog for analytics, flags or experiments | PostHog’s MCP server |
| You want your agent to explore data and write its own queries | PostHog’s MCP server |
| You want AI to watch recordings for you | PostHog Replay vision |
| You build with Claude Code and want one fix with the file and line, then a check | UserProof |
| You want a flat monthly price as traffic grows | UserProof |
Set up UserProof from the home page See pricing
Other agent-ready replay tools: LogRocket, Sentry, FullStory, Microsoft Clarity’s MCP server, and the full ranking of session replay tools for AI coding agents.
FAQ
What session replay tools does PostHog's MCP server have?
Eight: query-session-recordings-list, session-recording-get, session-recording-delete, session-recording-bulk-delete and four playlist tools. Replay vision adds 52 tools for AI scanners and observations (tools reference, read October 26, 2026).
Is PostHog's MCP server free?
Its docs say connecting and calling tools is free. Some tools use LLMs internally and may be billed as PostHog AI spend, and those need AI data processing turned on.
Can PostHog's MCP tell my agent what to fix?
Signals and Inbox group product signals into reports, and Inbox opens pull requests for actionable issues (first 3 a month free, then $15 each). UserProof's get_top_fix returns one fix ranked by your goal with the file and line, and check_fix tells you whether it moved the goal.
Can I use both?
Yes. They are separate MCP servers, and an agent can call both. Each needs its own snippet on your site.
Which coding agents work with each?
PostHog names Cursor, Claude Code, Claude Desktop, Codex, VS Code, Zed and Windsurf. UserProof documents Claude Code, Codex and Cursor.
Sources
- PostHog: PostHog MCP tools reference; PostHog MCP docs; PostHog replay over MCP; PostHog pricing (read October 26, 2026)
- UserProof: UserProof pricing; UserProof setup; plugin and MCP source (read October 26, 2026)
Corrections
If a fact here is out of date, the vendor’s own page is the source of truth. Report it through our GitHub issues. We re-read these pages every quarter.