By Ira Bodnar, Ryze AI · Updated · Connector tested November 13, 2026
Quick answer
The main way to connect Claude to PostHog is the Ryze AI connector. Create a PostHog personal API key, save it with your host on the PostHog card in Ryze AI, then add https://connector.get-ryze.ai/mcp as a custom connector in Claude, sign in to Ryze AI when Claude asks and start a new chat.
- In PostHog create a personal API key with the Query and Insight scopes
- In Ryze AI open Integrations, click Connect on PostHog, paste the key and host, pick the project
- In Claude open Customize, Connectors, then Add and Add custom connector, and paste the URL
- Click Connect if Claude shows it, pick the workspace, click Allow connection, open a new chat
Claude to PostHog at a glance
| Fact | Detail |
|---|---|
| Main way to connect | The Ryze AI connector: one hosted address, added in Claude as a custom connector or from the Ryze AI listing in the Connectors Directory |
| Connector address | https://connector.get-ryze.ai/mcp |
| What PostHog needs | A phx_ personal API key that can read the project and holds the Query and Insight scopes, plus your US, EU or self-hosted host. The key is saved in Ryze AI and never pasted into Claude |
| How Claude signs in | OAuth to Ryze AI. Claude calls the server from Anthropic's cloud, so claude.ai and Claude Desktop use the same steps |
| PostHog tools | 3, all on one project: posthog__read, posthog__write and posthog__delete |
| Can Claude change PostHog | Only when the key carries write scopes. Claude's Tool permissions can hold the write and delete tools on Needs approval or Blocked |
| Claude plan | Free, Pro, Max, Ultra, Team or Enterprise. Free allows one custom connector, and on Team and Enterprise an Owner adds it first |
| Ryze AI price | $89 per month for Paid Ads, after a 7-day free trial. Adding the connector costs nothing |
| PostHog's own option for Claude | The PostHog connector in Claude's Connectors Directory, labeled Anthropic verified, on PostHog's hosted server at https://mcp.posthog.com/mcp. The listing shows 57 tools |
| PostHog request limits | Per personal API key: 240 a minute and 1,200 an hour on analytics endpoints, 480 a minute and 4,800 an hour on other endpoints |
| Last tested | Connector address checked November 13, 2026. No step inside Claude and no PostHog call was run |
Before you start
- A Claude account. Anthropic lists custom connectors using remote MCP on Free, Pro, Max, Ultra, Team and Enterprise plans. Free users can add one custom connector. On Team plans only Owners and Primary Owners can add a custom connector; on Enterprise plans Owners, Primary Owners and people with a custom role that includes Manage access to Libraries can. Members then connect to it individually.
- A PostHog personal API key. It starts with phx_ and needs read access to the project, including the Query and Insight scopes. Add write scopes only if Claude should change objects. You also need the host: https://us.posthog.com for US Cloud, https://eu.posthog.com for EU Cloud, or your self-hosted URL.
- A Ryze AI workspace on a plan or free trial. Paid Ads is $89 per month with a 7-day free trial. The PostHog tools also work on SEO Starter ($129 per month, 3-day free trial) and Agency ($279 per month). Connecting the connector is free; without an active plan or trial, Claude can only list workspaces and connect platforms.
How to connect Claude to PostHog in claude.ai
Three steps happen in PostHog and Ryze AI, three in Claude. The connector URL is the same for every user.
Create a personal API key in PostHog
In PostHog go to Settings, Personal API keys and click Create key. Grant read access to the project with the Query and Insight scopes, add write scopes only for objects Claude should change, and copy the phx_ key. PostHog shows the value only once.
Connect PostHog in Ryze AI
In the Ryze AI app open Integrations, click Connect on the PostHog card, paste the key, enter the Host (US Cloud, EU Cloud or your self-hosted URL) and click Continue.
Pick the project
Choose the PostHog Project and click Save. One PostHog connection holds one project.
Add the custom connector in Claude
In claude.ai open Customize, Connectors (claude.ai/customize/connectors); in Claude Desktop, Customize sits in the sidebar. Click Add (some Anthropic pages print it as + Add), then Add custom connector; in some layouts you choose Custom and then Web instead. Enter the name Ryze AI and the MCP server URL https://connector.get-ryze.ai/mcp exactly, /mcp included. Keep the settings Claude marks Detected and do not choose No sign in: the Ryze AI server signs in with OAuth.
Connect and allow the workspace
Click Add. If the connector then shows Connect, click it. The Ryze AI page Connect your data sources to AI opens in the same tab on claude.ai and in your browser from Claude Desktop. Sign in (or click Sign up), choose the workspace that holds PostHog under Workspaces and click Allow connection. Ryze AI lets one connection cover up to 20 workspaces.
Open a new chat and ask your first question
Start a new chat. Connectors can be turned on or off per conversation from the + button, Connectors. Ask: List my PostHog dashboards and feature flags.
The Ryze AI PostHog guide covers the PostHog side of the connection, and the Claude connector FAQ covers the connection errors.
Sources: Anthropic: add a custom connector, Anthropic custom connectors, PostHog personal API keys, Ryze AI PostHog guide. Checked November 14, 2026.
Claude Desktop, Team plans and the Claude Directory
Anthropic says Claude reaches a remote MCP server from Anthropic's cloud, not from your device, on every Claude client including Desktop, Cowork and the mobile apps. So the steps are the same in Claude Desktop: open Customize, Connectors and add the URL. The other paths are below.
Team and Enterprise organizations
- An Owner or Primary Owner (on Enterprise, also a person with a custom role that includes Manage access to Libraries) opens Organization settings, Connectors and clicks Add.
- Choose Custom, then Web if Claude asks for the connector type. Enter the name Ryze AI and the URL https://connector.get-ryze.ai/mcp, and click Continue if the dialog shows it.
- Keep the settings marked Detected. Under Authentication choose Sign in now or Sign in when needed (not No sign in), then click Add.
- Each member opens Customize, Connectors and clicks Connect on Ryze AI. Anthropic shows a custom connector on a listed service's domain with that listing's name and branding, so it may not carry the Custom label; which one appears for this address was not tested.
- The member signs in on the Ryze AI page, picks workspaces, clicks Allow connection and starts a new chat.
Claude Directory listing
- Open the Ryze AI listing at claude.ai/directory/ryze-ai. It carries the Community label, and its notice says Community connectors get automated checks and are not verified by Anthropic.
- Sign in to Claude (when signed out the button reads Sign in to add) and add the connector from the listing.
- Complete the Ryze AI sign-in, choose workspaces, click Allow connection and start a new chat.
Claude and remote MCP servers: the facts in one table
| Topic | What Anthropic's pages say |
|---|---|
| Where you add it | Free, Pro, Max and Ultra: Customize > Connectors, then Add custom connector. Team and Enterprise: an Owner starts in Organization settings > Connectors, and each member connects afterwards |
| What the dialog asks for | A name and an HTTPS server address, then Authentication, the OAuth client and, in some organizations only, Request headers. Claude marks what it read from the server as Detected |
| Transport | Streamable HTTP for a remote address. The older HTTP+SSE transport still works while Anthropic phases it out |
| How sign-in starts | A 401 answer from the server. Claude then runs OAuth 2.0 with a PKCE S256 challenge and never sees your password |
| Network rules | Calls leave from Anthropic's cloud and need HTTPS on port 443. An address on another port does not connect |
| Asking before a tool runs | A prompt with Allow once and Always allow. Standing rules per tool or per group: Always allow, Needs approval or Blocked |
| Inside one conversation | The + button or typing /, then Connectors, switches a connector on or off. Tool access offers Auto, Always available and On demand |
| Size of a call | About 150,000 characters per tool result and 240 seconds per tool call |
| Local servers | Of the two, only Claude Desktop runs servers on your computer, as desktop extensions. claude.ai in a browser takes remote addresses only |
| Managing it later | Customize > Connectors, then the connector: Tool permissions, Disconnect, or Remove in the three-dot menu. Status reads Connected, Connect or Reconnect |
Sources: Anthropic: add a custom connector, Anthropic: connectors getting started, Anthropic: connector authentication, Anthropic: building connectors, Anthropic custom connectors. Checked November 14, 2026.
Claude Code, ChatGPT and Cursor
The same URL works in other MCP clients. PostHog is connected once in Ryze AI, so nothing changes on the PostHog side.
Claude Code
Run claude mcp add --transport http ryze https://connector.get-ryze.ai/mcp, then run /mcp inside Claude Code, select ryze and authenticate, and choose workspaces on the Ryze AI page. Claude Code needs a Pro, Max, Ultra, Team, Enterprise or Console account; the free claude.ai plan does not include it.
ChatGPT
OpenAI lists full custom MCP server support on ChatGPT Business and Enterprise/Edu, and read and fetch only for Pro. On the web open Plugins (chatgpt.com/plugins), select the plus button, choose Add custom MCP server, enter the name Ryze AI and the URL, set authentication to OAuth and finish the Ryze AI sign-in.
Cursor
Add the entry below to ~/.cursor/mcp.json (all projects) or .cursor/mcp.json (one project), then sign in on the Ryze AI page and click Allow connection.
{ "mcpServers": { "ryze": { "url": "https://connector.get-ryze.ai/mcp" } } }PostHog key scopes and permissions
PostHog access comes from the personal API key you paste into Ryze AI, so you set the limits when you create the key. Claude then adds its own per-tool approvals on top.
- Read scopes
- Create the key with read access to the project, including the Query and Insight scopes, as the Ryze AI setup asks. Those two cover HogQL queries and saved insights. Dashboards, feature flags, cohorts, persons and event definitions each have their own PostHog scope. Whether Ryze AI needs those scopes to list the objects was not tested, so add their read scopes if Claude should list them.
- Write scopes
- Add write scopes only for the objects you want Claude to change. What the agent can change is limited by the scopes on the key. A key with read scopes only keeps Claude read-only, because PostHog rejects any change.
- Host
- Enter https://us.posthog.com for US Cloud, https://eu.posthog.com for EU Cloud, or the URL of a self-hosted instance. Ryze AI works with all three.
- One project
- One connection holds one PostHog project. To switch, click View on the PostHog card in Ryze AI, disconnect with the trash icon and connect again. Another project can live in another Ryze AI workspace.
The PostHog scopes behind the three tools
You paste the PostHog key into Ryze AI during setup, and Claude signs in to Ryze AI with OAuth. PostHog gives a personal API key the same access as the person who created it, cut down to the scopes and projects chosen on the key. This table maps PostHog's scopes to the three tools.
| PostHog scope | What PostHog says it covers | With the Ryze AI connector |
|---|---|---|
| query:read | Running SQL and insight queries on project data, plus event, session and web vitals lookups. It has no write level | Named in the Ryze AI setup. Every HogQL question Claude asks runs on it |
| insight:read, insight:write | Insights with their queries, results and explanations | Read is named in the setup. Add write if Claude should save new or edited insights |
| dashboard:read, dashboard:write | Dashboards with their tiles, text cards and layout | Not named in the setup. Dashboard questions work when the key holds the read level. Not tested |
| feature_flag:read, feature_flag:write | Flags with release conditions, variants, payloads and rollout history | Not named in the setup. Read for questions about flags, write for Claude to change one. Not tested |
| cohort:read, cohort:write | Cohorts with filters, static member lists and calculation history | Not named in the setup. Read for listing cohorts, write for creating or deleting one. Not tested |
| person:read | Persons and their properties | Not named in the setup. posthog__read lists persons among the objects it reads. Not tested |
| event_definition:read | Event definitions, schemas and property groups | Not named in the setup. Questions about which events exist work when the key holds it. Not tested |
How PostHog applies scopes: each one has a read level to view data and a write level that can also create, change and delete it. A request that needs a scope the key lacks fails with a 403 naming that scope. A key can cover the whole organization, chosen projects or all projects, and its scopes can be edited later. Organization admins and owners can see each member's key that reaches the organization, with its scopes and last use.
Outside the key the Ryze AI setup describes: every write scope, a second project in the same workspace, and PostHog's MCP Server key preset, which PostHog made for its own server and CLI. Ryze AI documents its PostHog tools for insights, dashboards, feature flags, cohorts, persons, event definitions, projects and HogQL. The documentation does not list session recordings, experiments, surveys or error tracking, so do not count on Claude reaching those through Ryze AI.
Sources: PostHog API scopes, PostHog personal API keys, PostHog API overview. Checked November 14, 2026.
Removing access
To remove access, disconnect PostHog with the trash icon on the card in Ryze AI. To cut it off at the source, delete the personal API key in PostHog under Settings, Personal API keys. PostHog also scans GitHub for exposed phx_ keys and rolls them automatically (PostHog docs). In Claude, Disconnect on the connector's page signs Claude out and keeps Ryze AI in the list, while Remove in the three-dot menu takes it away.
What Claude can do in PostHog
The Ryze AI connector exposes three PostHog tools, one each for reading, writing and deleting. They act on the single project you picked.
| Tool | Type | What it does |
|---|---|---|
| posthog__read | Read | Reads projects, insights, dashboards, feature flags, cohorts, persons and event definitions for the connected project, and runs HogQL analytics queries. |
| posthog__write | Write | Creates or updates PostHog objects such as insights, dashboards, feature flags and cohorts. |
| posthog__delete | Delete | Deletes a PostHog object by path, for example a cohort. The deletion is irreversible. |
Claude can ask before a tool call, with Allow once and Always allow. For a standing rule, open Customize, Connectors, Ryze AI, Tool permissions and put the write and delete tools on Needs approval, or on Blocked if Claude should never change the project; the read tool can stay on Always allow. During Research Claude may call connector tools without asking again, and Anthropic advises turning write tools off before such a run. Scheduled tasks run without an approval step, and the Ryze AI chat approvals picker starts on Skip all approvals, so the key scopes are the stronger guard. Ryze AI also bounds every HogQL query: give it a time window and at most 1,000 rows.
PostHog limits in numbers
PostHog publishes these figures, except the HogQL row, which Ryze AI sets, and the last row, which comes from Anthropic. None was measured for this guide. PostHog counts most of its rate limits per personal API key, here the key saved in Ryze AI.
| Limit | Number |
|---|---|
| Analytics endpoints | 240 requests per minute and 1,200 per hour for each personal API key. This covers calculating insights and retrieving persons. |
| Other create, read, update and delete calls | 480 per minute and 4,800 per hour for each key. |
| Query endpoint | 2,400 per hour on the API overview. The queries page sets it per project, with 240 per minute, 3 queries at once, 60 threads per query and 10 seconds of execution time. Some older accounts have 120 per hour. |
| Hourly read budget | Each project has a budget of bytes read per hour for personal API key queries. PostHog publishes no size. Past it, the answer is 429 with the code api_queries_budget_exceeded and a Retry-After header; cached results still return. |
| Rows per query | 100 unless the query sets its own LIMIT, and up to 50,000 with one. OFFSET gets HTTP 400 for a personal API key, so page by timestamp. |
| HogQL through Ryze AI | A time window and 1,000 rows at most. Ryze AI sets this bound; PostHog's own row figures are in the row above. |
| List pages | Usually 100 results per page, sometimes 500 or 1,000, with next and previous links. |
| Keys per person | Up to 10 personal API keys. |
| Dynamic cohorts | Recalculated once every 24 hours. |
| On Claude's side | Anthropic cuts a tool result off at about 150,000 characters and a tool call at 240 seconds. |
Sources: PostHog API overview, PostHog queries, PostHog personal API keys, PostHog cohorts, Anthropic: building connectors. Checked November 14, 2026.
First 3 prompts to run
These prompts describe what to ask. The answers come from your own PostHog project, so this page does not show sample numbers.
1. Inventory the project
List my PostHog dashboards, feature flags and cohorts, and tell me which dashboards have no insights on them.
Tools used: posthog__read
What comes back: A list of dashboards, flags and cohorts for the connected project, grouped by type. If the key lacks the read scope for a type, PostHog answers with a 403 that names it.
2. Run a HogQL query
Run a HogQL query for the number of $pageview events per day over the last 14 days, limited to 1,000 rows, and give me the SQL you used.
Tools used: posthog__read
What comes back: One row per day with the event count, plus the query text so you can re-run it in PostHog.
3. Create an insight
Create a PostHog insight that counts unique users who triggered $pageview each week, name it Weekly visitors from Claude, and add it to a new dashboard.
Tools used: posthog__write
What comes back: The new insight and dashboard, created in your project. The key needs insight:write and dashboard:write. Set write tools to Needs approval in Claude first if you want to confirm the change.
Event names such as $pageview are PostHog defaults; use your own event names where they differ.
10 jobs to hand to Claude
Ten jobs you can give Claude once the connector is on, each beside the tool it calls. The default key is the one the setup describes: read access to the project with the Query and Insight scopes and no write scope. None of the ten was run for this guide.
| Job | Ask Claude | Tool | Works on the default key |
|---|---|---|---|
| Daily pageviews | Write a HogQL query for pageview events per day over the last 7 days, cap it at 1,000 rows and show me the SQL. | posthog__read | Yes, through query:read |
| Traffic by campaign | Split last month's sessions by UTM source with a HogQL query over sessions. | posthog__read | Yes, through query:read |
| Funnel drop-off | Open my saved signup funnel insight and give me the number of people at each step. | posthog__read | Yes, through insight:read and query:read |
| Dashboard names | List this project's dashboards by name so I can point you at one later. | posthog__read | Only if the key also holds dashboard:read. Not tested |
| Flag check before a launch | Show every feature flag with its current rollout settings. | posthog__read | Only if the key also holds feature_flag:read. Not tested |
| Cohort definitions | List the cohorts here and what each one is defined by. | posthog__read | Only if the key also holds cohort:read. Not tested |
| Event inventory | Which event definitions exist in this project? I want to check before building a report. | posthog__read | Only if the key also holds event_definition:read. Not tested |
| Campaign dashboard | Build a trends insight for landing page signups and place it on a new dashboard for the campaign. | posthog__write | No. The key needs insight:write and dashboard:write |
| Campaign cohort | Make a cohort of the people who arrived from one campaign. | posthog__write | No. The key needs cohort:write |
| Remove a test cohort | Delete the old test cohort I name. I know this cannot be undone. | posthog__delete | No. The key needs cohort:write |
With the write and delete tools on Needs approval in Claude, the last three jobs wait for your click even when the key allows them.
7 ways to connect Claude to PostHog
Seven ways put PostHog data in front of Claude in claude.ai or Claude Desktop, and in each case the vendor names Claude. This guide treats the first as the main way: the Ryze AI connector keeps the PostHog key inside Ryze AI, and Claude needs only one address and an OAuth sign-in, the same on the web and in the desktop app. The other six were read on their own pages on November 14, 2026. None of them was run for this guide.
| Way | Runs where | Setup in Claude | What Claude reaches in PostHog | Price |
|---|---|---|---|---|
| Ryze AI connector (the main way) | Hosted by Ryze AI, one address for every user | Customize, Connectors, Add custom connector with https://connector.get-ryze.ai/mcp, or the Ryze AI listing in the Connectors Directory, which has the Community label. OAuth sign-in to Ryze AI | One project: insights, dashboards, feature flags, cohorts, persons, event definitions and HogQL queries. Changes to insights, dashboards, flags and cohorts when the key has write scopes | $89 a month on Paid Ads, 7-day free trial first |
| PostHog connector in Claude's directory | PostHog's hosted server at https://mcp.posthog.com/mcp, which sends the sign-in to your US or EU region | Open PostHog in the Connectors Directory, add it and sign in to PostHog. The listing is labeled Anthropic verified. If an admin added it for the team, each member still connects their own account | 57 tools on the listing, reading and writing, such as query-run, dashboard-create and update-feature-flag. PostHog's page counts 1,284 tools in 78 categories on the server | No price on the listing. PostHog says connecting and calling tools is free, while some tools use LLMs and may be billed as PostHog AI spend |
| PostHog AI plugin, in Cowork | Plugin files, with PostHog's hosted server behind them | PostHog's README for Claude Cowork: Settings, Plugins, Add, Add marketplace, enter PostHog/ai-plugin, Sync, then Browse and install PostHog. Anthropic says a plugin does not sign in to its connectors; you connect each one from the plugin's Connectors tab | The hosted server's tools, plus 30+ PostHog skills and slash commands such as /posthog:insights | No price stated. MIT licence |
| Composio | Hosted by Composio | Add https://connect.composio.dev/mcp as a custom connector, authenticate with Composio and allow access. PostHog is linked in Composio with your own API key. Composio's list names Claude Cowork, and its intro names Claude in general terms | 686 tools and 11 triggers in its Posthog toolkit | The free plan stops app tool calls at 10,000 a month. Standard is $18 per seat per month |
| Windsor.ai | Hosted data connector, with Windsor MCP as a destination | Add PostHog as a data source at onboard.windsor.ai, then follow Windsor's own Claude setup page, which was not opened for this guide | 101 PostHog fields, 6 metrics and 95 dimensions, from insights, events, cohorts, feature flags, persons and more. PostHog is not among the platforms Windsor names for write-back | Forever Free after a 30-day trial: 1 data source and the past 30 days of data. Basic shows $23 per month, or $19 on annual billing |
| Coupler.io | Hosted data export service | Enter a PostHog private API key and your instance domain, choose a data entity and a start date, then pick Claude as the AI destination | Scheduled exports of 7 entities: projects, cohorts, feature flags, persons, events, annotations and insights | Free plan with 1 data source and manual refresh. Starter is $32 per month, or $24 billed annually, with 7 free days to start |
| AnythingMCP, open source | AnythingMCP Cloud in Frankfurt, or your own Docker host | Give AnythingMCP a phx_ key and the PostHog host, then add your AnythingMCP server address to Claude with OAuth. Its PostHog guide links a setup page for Claude | 28 tools, such as posthog_run_hogql_query. Three delete tools start switched off, and the guide says it does not create or edit dashboards, cohorts or experiments | Cloud Starter is 19 euros per month after a 7-day trial. Self-hosted Community edition: free under AGPL-3.0 for up to 3 active users |
PostHog has two more routes to the same hosted server for Claude. Claude Desktop users can run npx @posthog/wizard mcp add, which PostHog says writes the server into Claude Desktop, among other clients. On Team and Enterprise, an Owner can set up PostHog's managed authorization, so members reach PostHog when they sign in to Claude, with no PostHog sign-in or key of their own. PostHog asks for its Scale or Enterprise add-on, a verified domain, members under the same email and at least the user:read, organization:read and project:read scopes; without the add-on it refuses token requests with access_denied. For Claude Code, PostHog gives the command claude mcp add --transport http posthog https://mcp.posthog.com/mcp -s user. Its docs also cover Cursor, Codex, VS Code, Windsurf and Zed, and name Lovable, Replit and v0 as platforms.
Options that do not fit claude.ai or Claude Desktop
- PostHog plugin for ChatGPT. It sits in ChatGPT's plugin directory and names ChatGPT only.
- PostHog CLI. Built for terminals, scripts and CI. PostHog names Claude Code, Codex and other coding agents for it, and its CLI page points an agent without terminal or shell access to MCP.
- Zapier MCP. Zapier lists Claude among its MCP clients, yet its PostHog page has one action, Capture Event, and no action that reads insights, dashboards, flags or cohorts.
- A PostHog key in a request header. PostHog offers a personal API key sent as a bearer token for clients without OAuth. Claude completes PostHog's OAuth sign-in, and Claude's Request headers section is a beta that only a limited set of organizations sees.
- Retired or unrelated packages. posthog-for-claude, an earlier Claude Code plugin, and the standalone PostHog/mcp repository are archived. @posthog/agent-toolkit is deprecated in favor of the MCP server, and the npm package @posthog/mcp is an SDK for tracking tool usage in your own MCP server.
Which one to pick
Ryze AI is the best way to connect Claude to PostHog when you want one project's insights, flags, cohorts and HogQL answers on the same address as your ads and SEO platforms while the PostHog key stays in Ryze AI, with one caveat: a workspace holds one project, and the three Ryze AI tools do not list session recordings, experiments, surveys or error tracking. PostHog's own connector reaches those areas, carries the Anthropic verified label and needs no Ryze AI plan, so it suits a chat where PostHog is the only tool. Add the PostHog AI plugin in Cowork if you want its skills. Composio fits a team that already routes other apps through it. Windsor.ai and Coupler.io fit reporting on fixed fields or scheduled exports, and AnythingMCP fits a team that wants to host the server itself.
See the full Ryze AI tool list on the PostHog MCP page.
Sources: PostHog in Claude's Connectors Directory, PostHog MCP server, PostHog MCP for Claude Desktop, PostHog MCP for Claude Code, PostHog managed authorization for Claude, posthog.com/mcp, PostHog AI plugin, PostHog CLI, Anthropic plugins overview, Composio Posthog toolkit, Composio pricing, Windsor.ai PostHog fields, Windsor.ai pricing, Coupler.io PostHog source, Coupler.io pricing, AnythingMCP PostHog guide, AnythingMCP pricing, Zapier PostHog MCP, Anthropic custom connectors, Ryze AI PostHog guide, Ryze AI Claude connect guide. Checked November 14, 2026. The Ryze AI Claude pages were last checked November 8, 2026.
Claude, PostHog and MCP: dates that matter
| Date | What happened |
|---|---|
| November 25, 2024 | Anthropic open-sourced the Model Context Protocol. Local MCP servers worked in the Claude Desktop apps on every plan. |
| May 1, 2025 | Anthropic launched Integrations, its first support for remote MCP servers on the web and in the desktop app, in beta on Max, Team and Enterprise. |
| May 30, 2025 | A PostHog blog post linked the docs of its MCP server. It is the earliest dated PostHog page about the server found for this guide; PostHog gives no launch date. |
| June 3, 2025 | Integrations opened to the Pro plan. |
| July 14, 2025 | Anthropic opened its connectors directory on the web and in the desktop app. Remote connectors were for paid plans only at that point. |
| February 25, 2026 | Claude Desktop gained a Customize section that holds skills, plugins and connectors in one place. |
| February 27, 2026 | PostHog's changelog announced the PostHog connector for Claude, and on the same day cohort management through its MCP server. |
| August 2026 | The month the Ryze AI listing in Claude's Connectors Directory gives as its Added date. |
| September 16, 2026 | Anthropic merged Cowork and chat into one Claude experience, rolling out to Pro and Max. Connectors carry over. |
| October 1, 2026 | PostHog's MCP server began keeping clients with non-rotating refresh tokens signed in, where they used to need a new authorization after roughly 67 days. |
| October 22, 2026 | PostHog's wizard began detecting Claude Code and Codex and installing MCP for them, with a separate flow for browser and account-based clients such as Claude. |
| November 5, 2026 | PostHog's new events retention periods apply to organizations that start a paid plan on or after this day. PostHog lists 1 year on Free and 2 years on pay-as-you-go. |
| November 14, 2026 | posthog.com/mcp counted 1,284 tools in 78 categories, and the Ryze AI listing in Claude's directory still showed the Community label. |
Sources: Anthropic: Model Context Protocol, Anthropic: Integrations, Anthropic: connectors directory, Claude release notes, PostHog blog, May 30, 2025, PostHog changelog, PostHog events retention, posthog.com/mcp, Ryze AI listing in the Claude Directory. Checked November 14, 2026.
How this guide was tested
One check was run for this guide, on November 13, 2026 at about 04:20 UTC: the live connector address and its two discovery files. Everything else here comes from the Ryze AI help center and app, and from PostHog's, Anthropic's and the other vendors' pages, read on November 14, 2026.
| What was checked | What came back |
|---|---|
| A request to https://connector.get-ryze.ai/mcp with no sign-in | HTTP 401 with a WWW-Authenticate header that points to the sign-in details. Anthropic's docs say a 401 is the answer that makes Claude start its OAuth sign-in. |
| The connector's two discovery files | Both loaded. They name the authorize, token and register endpoints, the authorization code and refresh token grants, PKCE with S256, and the scopes mcp:read and mcp:write. Claude sends an S256 challenge on every authorization request, so the two match. |
| Calls to PostHog | None. No PostHog test account was connected, so no PostHog call was run. |
| The three PostHog tool names | Taken from the Ryze AI app source. They were not seen in a live tool list. |
| The steps inside Claude | None run. No connector was added in claude.ai or Claude Desktop. |
Not run for this page: any step in claude.ai or Claude Desktop, and any call to PostHog. Nothing was added, signed in to or called in Claude, so the Claude steps rest on Anthropic's pages as they read on November 14, 2026. Which OAuth client option Claude marks Detected for the Ryze AI address, and how the Team and Enterprise list labels it, were not checked. With no PostHog account connected, this guide has no table of tested calls and no error messages copied from a test. The errors section below quotes PostHog's documented answers instead.
PostHog errors and fixes
- PostHog tools do not appear in Claude
- The Ryze AI help center says Claude keeps the Ryze AI tool list from the moment you add the connector. If PostHog was connected afterwards, remove the Ryze AI connector, add it again and start a new chat.
- Claude says a subscription is required
- The PostHog tools need an active Ryze AI plan or free trial. Without one the connector can only list workspaces and connect platforms.
- PostHog rejects a change
- The personal API key has read scopes only, or lacks the scope for that object. PostHog's docs say such a request fails with a 403 that names the missing scope. Create a key with the write scopes you want and reconnect PostHog in Ryze AI.
- A query is refused or stops
- Through Ryze AI a HogQL query needs a time window and a row limit of 1,000 at most, so ask Claude for both. PostHog also ends a query after 10 seconds of execution time and answers HTTP 400 when a personal API key uses OFFSET; narrow the dates or page by timestamp.
- Queries are rate limited
- PostHog allows 240 requests per minute and 1,200 per hour on analytics endpoints for each personal API key. The query endpoint allows 2,400 per hour, which the queries page sets per project, with 240 per minute. When the project's hourly read budget runs out, PostHog answers 429 with api_queries_budget_exceeded and a Retry-After header. Wait and retry, or spread the questions out (PostHog docs).
- The wrong project is connected
- PostHog holds one project per connection. Click View on the PostHog card in Ryze AI, disconnect with the trash icon, connect again with the right key and pick the project.
- The one-time connection link is refused
- If you asked the Ryze AI agent in Claude to connect PostHog, the link is valid once and for 24 hours, and it is refused while PostHog is already connected in the workspace. Change keys or projects in the Ryze AI app.
- The EU or self-hosted project fails to connect
- Check the Host field. Use https://eu.posthog.com for EU Cloud and your own URL for self-hosted, not the US address. If the key itself is wrong, PostHog's docs describe the answer as a 401 with the message Invalid Personal API key.
- Claude does not offer Ryze AI in a chat
- Turn the connector on from the + button, Connectors, and start a new chat after connecting.
Sources: PostHog API overview, PostHog queries, PostHog API scopes, Ryze AI Claude connector FAQ. Checked November 14, 2026. The Ryze AI Claude pages were last checked November 8, 2026.
Terms on this page
- MCP
- The Model Context Protocol, which Anthropic open-sourced in November 2024. Claude uses it to call tools on remote servers such as the Ryze AI connector.
- Custom connector
- A remote MCP server you add to Claude by pasting its address. Anthropic does not review it, and Claude never suggests it in a chat.
- Verified and Community
- Labels in Claude's Connectors Directory. Verified means Anthropic tested the tools for quality and compatibility. Community means a third party built it and Anthropic screened it without an in-depth review. PostHog's listing is Verified; the Ryze AI listing is Community.
- Personal API key
- PostHog's key for one person. It starts with phx_, carries the scopes chosen for it, and is pasted into Ryze AI during setup.
- Scope
- One permission on a PostHog key for one area, such as insights or feature flags, with a read level and a write level.
- HogQL
- PostHog's SQL, a wrapper around ClickHouse SQL. PostHog uses the name internally for SQL access.
- Insight
- The basic building block of PostHog product analytics: a saved trend, funnel, retention chart or similar.
Sources: Anthropic: Model Context Protocol, Anthropic: directory and custom connectors, Anthropic: connector labels, PostHog API scopes, PostHog SQL, PostHog insights. Checked November 14, 2026.
Frequently asked questions
How do I connect Claude to PostHog?
Create a PostHog personal API key with the Query and Insight scopes, open Integrations in Ryze AI, click Connect on the PostHog card, paste the key, enter the host and pick the project. Then in claude.ai open Customize, Connectors, Add custom connector, enter the name Ryze AI and the URL https://connector.get-ryze.ai/mcp, click Connect if it shows, allow the workspace and start a new chat.
Does PostHog have its own MCP server for Claude?
Yes. PostHog hosts a free MCP server at https://mcp.posthog.com/mcp and lists it in Claude's Connectors Directory as the PostHog connector, labeled Anthropic verified. Sign-in goes through PostHog and routes to your US or EU region, and PostHog's docs also name Claude Code and Claude Desktop as supported clients. The Ryze AI connector is a separate path with three PostHog tools.
Is PostHog's connector in Claude's directory the same as the Ryze AI connector?
No, they are two separate servers. PostHog's connector reaches PostHog's hosted server with a PostHog sign-in, and its listing shows 57 tools. The Ryze AI connector at https://connector.get-ryze.ai/mcp reaches PostHog with the personal API key saved in Ryze AI and has three PostHog tools.
Which PostHog connector should I use with Claude?
If PostHog is the only tool you need in the chat, PostHog's own server lists more areas, such as error tracking and experiments. Ryze AI is the best PostHog connector for Claude when you want one connector with a Ryze AI plan behind it and a three-tool read, write and delete setup, but it covers fewer PostHog areas than PostHog's own server.
Is the Ryze AI connector verified by Anthropic?
No. The Ryze AI listing in Claude's Connectors Directory carries the Community label: a third party built it, and Anthropic screened it before listing without an in-depth review. Directory and custom connectors share the same runtime, sign-in and tool calls in Claude.
What PostHog access do I need?
A personal API key that starts with phx_, with read access to the project including the Query and Insight scopes, and the host for your region. Add the read scopes for dashboards, flags or cohorts if Claude should list them, and write scopes only if Claude should change objects.
Which Claude plan do I need?
Anthropic lists custom connectors on Free, Pro, Max, Ultra, Team and Enterprise plans. Free users can add one custom connector. On Team and Enterprise an Owner adds the Ryze AI connector first, then each member clicks Connect.
Can Claude change or delete things in PostHog?
Yes, when you ask. posthog__write creates or updates insights, dashboards, feature flags and cohorts, and posthog__delete deletes an object by path, which is irreversible. A key with read scopes only keeps Claude read-only.
Can I require approval before Claude changes PostHog?
Yes. In Claude open Customize, Connectors, Ryze AI, Tool permissions and set the write and delete tools to Needs approval, or to Blocked. Scheduled tasks run without an approval step.
Do the PostHog tools work in the Claude mobile app?
Anthropic says yes: a remote connector added on the web or in Claude Desktop works in the mobile apps after your next sign-in there. This was not tested with Ryze AI. Adding a connector from a phone is in beta, so add it on a computer. Research started in the mobile app searches the web only and skips connectors.
How many requests can Claude send to PostHog?
PostHog counts most limits per personal API key. Analytics endpoints take 240 requests a minute and 1,200 an hour, and other endpoints take 480 a minute and 4,800 an hour. Queries also draw on an hourly read budget per project, and Claude allows up to 240 seconds for one tool call.
Does Claude work with PostHog EU Cloud or self-hosted?
Yes. Enter https://eu.posthog.com for EU Cloud, https://us.posthog.com for US Cloud, or your self-hosted URL as the host when you connect PostHog in Ryze AI.
Can I connect more than one PostHog project?
Not in one workspace. A PostHog connection holds one project. Another project can live in another Ryze AI workspace, and one Claude connection can cover up to 20 workspaces.
Does it cost anything?
Connecting the connector is free. The PostHog tools need an active Ryze AI plan or trial: Paid Ads is $89 per month with a 7-day free trial, SEO Starter is $129 per month with a 3-day free trial and Agency is $279 per month. Annual billing takes 30% off the first year.
Why do the PostHog tools not show up in Claude?
According to the Ryze AI help center, Claude keeps the tool list from when you added the connector. Remove the Ryze AI connector, add it again and start a new chat.