by Typesense
Updated October 19, 2026
MCP100 verdict
Not currently recommended. Shortlist — editorially admitted, not a Recommended endorsement.
MCP100 Assessment
Editorial verdict
Reviewed October 19, 2026
Task Fit
Mapped
Specialized workflows
Trust
Status warning
Prefer Search only permissions; key auth does not enable full data tools
Maintenance
Date unavailable
No dated repository or package update is currently stored.
Popularity
Signals unavailable
No independently comparable adoption metric is currently stored.
Evidence scope
No MCP-specific repository metric is stored
No MCP-specific package metric is stored
Stars observed October 20, 2026 · Downloads observed October 20, 2026
Typesense Cloud MCP is a hosted service. No current first-party public repository containing the hosted server implementation was verified.
Typesense Cloud MCP is hosted; no exact public installable MCP package was verified.
Publisher activity observed October 19, 2026
Current source snapshot
Source relationship: First-party MCP
Reviewed October 19, 2026
Evidence-backed setup
Choose your MCP client. A client connects to the MCP server; the language model is selected inside that client. Only stored, source-backed configurations are shown.
Transport not recorded
Complete client-specific setup, verification and removal are not documented in this reviewed source.
Publisher-documented setup is not the same as MCP100 execution testing. “MCP100 logic checked” means the stored configuration shape was reviewed, not that third-party MCP software was run.
Evidence checked 2026-10-19
Type-appropriate, source-backed setup for this mcp server profile.
{"url":"https://cloud.typesense.org/mcp/v1"}Capability
Typesense Cloud MCP is a Typesense MCP integration for typesense cloud clusters, collections, search indexing and monitoring. Use for teams operating hosted Typesense Search. Cluster provisioning, schema changes and indexing can cost money or change production data.
Permissions determine tool surface; preset and OAuth capability dependent
Not measured yet
Estimated token footprint of the MCP tool schemas exposed to the agent.
https://cloud.typesense.org/mcp/v1
Reviewed access
Related reviewed MCPs
Let AI clients inspect and automate web browsers with Playwright.
Choose this instead if
Choose Playwright MCP instead if inspecting and testing web interfaces through accessibility snapshots.
Let AI clients inspect and automate web browsers with Playwright.
Use Stripe's remote MCP endpoint for supported payment workflows.
Choose this instead if
Choose Stripe MCP instead if searching Stripe documentation and support knowledge from an MCP client.
Use Stripe's remote MCP endpoint for supported payment workflows.
Manage and search Redis data through Redis's MCP server.
Choose this instead if
Choose Redis MCP instead if inspecting and manipulating Redis strings, hashes, lists, sets, sorted sets, streams, JSON, and search indexes.
Manage and search Redis data through Redis's MCP server.
Typesense is listed as First-party MCP in MCP100’s current evidence.
Current official resources:
Original product, capability, tool, and usage documentation from the MCP publisher. Installation sections appear under Install & Agents.
Title: Typesense Cloud MCP Server | Typesense
URL Source: https://typesense.org/docs/guide/typesense-cloud/mcp-server.html
Markdown Content: Connect your AI agent to Typesense Cloud and it can create and configure clusters, build collections, index your data, tune search and read metrics, all from the conversation. You decide what it is allowed to do when you connect it, and you can disconnect it at any time.
Add the Typesense Cloud MCP server to your AI Agent:
https://cloud.typesense.org/mcp/v1
It is a remote MCP server over Streamable HTTP with OAuth, so any client that supports those works.
Here are instructions for common agents:
Claude Code:
shell
claude mcp add --transport http typesense-cloud https://cloud.typesense.org/mcp/v1
Then open a new Claude Code session, type /mcp, select the typesense-cloud MCP you just added, and select "Authenticate"
Codex CLI:
shell
codex mcp add typesense-cloud --url https://cloud.typesense.org/mcp/v1
Codex reads your project's AGENTS.md every turn. Add this line to it so Codex reaches for Typesense Cloud instead of asking you for keys:
plaintext
For anything about my Typesense Cloud account, clusters or collections, use the typesense-cloud MCP tools.
Cursor (mcp.json, MCP setup docs):
json
{ "mcpServers": { "typesense-cloud": { "url": "https://cloud.typesense.org/mcp/v1" } } }
claude.ai and Claude Desktop (custom connector docs):
Go to Customize, then Connectors, then Add custom connector, and paste this URL: https://cloud.typesense.org/mcp/v1
ChatGPT (MCP setup docs):
In the ChatGPT desktop app, go to Settings, then Plugins, then MCP, then Add, then Add MCP Server. Name it "Typesense Cloud", set the type to Streamable HTTP and the URL to https://cloud.typesense.org/mcp/v1, and save.
Then ask: "Using Typesense Cloud MCP, show me my clusters."
Other Clients:
Any other client that supports MCP over Streamable HTTP with OAuth works the same way.
Once you add the MCP Server, follow your client's instructions to authenticate the MCP Server.
Your browser will open a Typesense Cloud consent page.
Sign in (or sign up), choose the account the agent should work on, choose what permissions it has, and click Connect.
To let your agent try Typesense Cloud before you sign up, click "Try it without an account" instead. See Starting without an account.
If you're part of a team account, the permissions in the consent screen will be capped at permissions your RBAC role in your team account already gives you. So for eg: someone on your team with a curator role, connecting their agent, will only be able to add permissions for searching and curation, and not any of the other permissions.
TIP
Once connected, paste this into your agent:
plaintext
Using MCP, set up Typesense Cloud for me: create a cluster, then index my data (it's in <describe or point to it>) and build a first search.
The agent creates a free-tier cluster, asks what your data looks like, proposes a schema, imports the data and runs a search you can check. From there, ask your agent for what you need. Here are some examples for inspiration:
Working with your data:
products.jsonl"products collection and a few sample documents"brand as a facet field to the products collection"products_live alias at the products_v2 collection"orders where status:=cancelled"orders collection to a file"Searching:
products and articles for trail running in one call and show me the top three of each"articles for getting started with vector search"brand and show me two per brand"Tuning search results:
created_at descending"sneakers"sneakers also matches trainers"title but none on sku"the, a and an as stopwords for the articles collection"storefront"products so people can type cheap red shoes for kids"docs collection"Search Typesense docs:
The MCP server will automatically query the docs for the version your Typesense Cloud cluster is running on
drop_tokens_threshold do and what should I set it to?"Managing your cluster's infrastructure:
max-per-page to 500 on this cluster"staging-old cluster"Analyzing your billing data:
Your agent can create a cluster before you have a Typesense Cloud account, and you claim the cluster into your account later.
When your client connects, click "Try it without an account" on the Typesense Cloud sign-in page. The agent sees all of its tools, but until you connect an account only three of them work: creating a free cluster, checking on it and searching the Typesense docs. The other tools tell the agent that you need to connect an account first.
The cluster is the free-tier shape (0.5 GB memory, 2 vCPU burst, one node), in the region nearest the agent unless it asks for another one. The agent gets the cluster's hostname and an API key, and works with it directly through the Typesense Server API. That key can create collections, import, search and set up synonyms, curation, presets, stopwords and analytics rules. It can't delete anything or manage API keys.
The agent also gives you a claim link. Open it, sign in or sign up and pick the account that should own the cluster: your own or one of your team accounts. The cluster keeps its ID, hostname and data, so nothing changes for your app.
If you open the claim link in the browser where you clicked "Try it without an account", Typesense Cloud also offers to connect the agent to that account, and you pick its permissions. The agent works on your account from its next request, so you don't need to reconnect it from your client.
To connect the agent to your account without claiming a cluster, ask it for its connect link. That link only works in the browser where you clicked "Try it without an account".
Note
An unclaimed cluster is deleted 72 hours after it was created. A connection without an account expires after 7 days, unless you connect it to an account.
TIP
When you connect your agent, the permissions page shows five presets:
| Preset | The agent can |
|---|---|
| Search only | Search collections and read schemas, list clusters. |
| Search and curate | Search, plus synonyms, curation, presets and stopwords, plus list clusters. |
| Manage clusters | Create, resize, rename, restart and tune clusters, and read their backups. No access to the data inside them. |
| Build (default) | Manage clusters and everything inside them. No terminating, cloning or billing. The only key it can create is a search-only cluster key for an app. |
| Everything | Every capability your role allows, including terminating clusters. |
You can also enable or disable individual permissions. Cluster permissions are the same ones a Cluster Management API key can carry:
clusters:read: list clusters and read their configuration, status, history and server parametersmetrics:read: read cluster metrics, request logs and each node's healthsearch_insights:read: read a collection's search insights (its top no-hit and popular queries, top facet values, field names and synonym sets) and its synonym suggestionsclusters:create: create clusters (billed to the account)clusters:configure: resize, change topology, rename and toggle auto-upgrade (billable changes)clusters:restart: start and cancel rolling restartscluster_parameters:manage: change Typesense server parameters and synonym suggestion settings, run or dismiss synonym suggestions (suggestion runs use your own LLM provider key and are billed by that provider) and read and change analytics rule notifications (scheduling them also needs search_insights:read)clusters:clone: clone clusterscluster_credentials:issue: issue a cluster's admin and search-only keyscluster_credentials:manage: list, create and revoke cluster API keyscluster_network:manage: read and replace a cluster's IP allowlist (which IPv4 addresses can reach it)backups:read: list a cluster's backups and read its backup settings and their estimated costbackups:manage: turn backups on or off, change their schedule and regions (billed) and delete backupsbackups:restore: restore a backup to a new cluster (billed), or onto the backed-up cluster, replacing all of its databilling:read: read invoices, the billing summary (prepaid credit balance, spend by month, the upcoming invoice and the billing cycle), invoiced usage, what each cluster cost per month and pricesbilling:manage: switch the account between weekly and monthly billingclusters:terminate: terminate clustersData permissions match the cluster UI dashboard's permissions:
data:view: read collections, schemas, documents and aliasesdata:search: search, and create a search-only cluster key for an app (if your role manages keys)data:full_access: manage collections (create, change and drop), documents (add, update and delete) and aliases. It includes data:view and data:search. It doesn't include the data:manage_* permissions belowdata:manage_synonyms (also needed to accept a synonym suggestion), data:manage_overrides (curation), data:manage_presets, data:manage_stopwords, data:manage_stemmingdata:manage_analytics_rules, data:manage_conversation_models, data:manage_natural_language_modelsEvery tool is offered to every connection, but each tool call is checked against the permissions you granted. When a call is refused, the agent is told which permission it's missing, along with a link to pass on to you. The link opens "Edit capabilities" for that connection with the permission already ticked. After you click "Save", the agent has that permission on its next call. You can also change a connection's permissions yourself by going to Account > Connected apps and clicking on "Edit capabilities". Only the person who connected an agent can change what it can do.
The tools that change things (manage_cluster, documents and change_cluster_data) are marked destructive, so clients that honor that annotation ask you before running them.
On a team account you can only grant what your own role allows. Options your role doesn't cover are grayed out with the reason.
These are the tools available for your agent to use once it is connected. Agents read the tool descriptions, so you don't need to name tools or actions in your prompts.
whoami: which account and permissions the connection has, the clusters on the account, and what to do with them first. Agents call it first.
search_typesense_docs: searches this documentation, matched to your cluster's Typesense version.
search_documents: searches a collection, including multi-search and vector or hybrid queries.
get_clusters: reads everything about your clusters and account:
show_cluster: shows a cluster's status, configuration, any change in progress and its last hour of metrics, in one call. Chat apps that support panels show it as a cluster panel.
manage_cluster: makes changes:
documents: indexes, updates, fetches, deletes, imports and exports documents.
read_cluster_data and change_cluster_data: everything else in the Typesense Server API by path: collections, aliases, synonyms, curation, presets, stopwords, stemming dictionaries, analytics, conversation models and natural language search models. This also covers the cluster's own /health and /metrics.json.
submit_feedback: lets the agent tell the Typesense Cloud team about a paper cut it hit, such as a tool that refused something that should work or a missing action. The team reads every submission, but nobody replies through the tool. For anything blocking you, contact support.
Typesense Cloud performs data operations against your cluster with a key it keeps server-side. The agent does not receive your cluster's admin key.
Chat apps that support MCP Apps, such as claude.ai, ChatGPT and Cursor, show some results as interactive panels in the conversation. Other clients get the same results as text.
show_cluster that tracks the change until it's done.A search from the panel needs data:search, and the cluster panel shows metrics only when the connection has metrics:read.
The agent imports up to 1,000 documents or 10 MB in one call. For anything larger, and for every export, it receives a curl command to run on your machine instead, so the data moves directly between your machine and the cluster. The command uses a temporary key that Typesense Cloud creates for that one job: scoped to that collection and that action, valid for 24 hours, fetched once into a private file on your machine.
Browser-based assistants such as claude.ai and ChatGPT cannot run commands, so their imports stay within the inline limit.
A new account can create a free-tier cluster (0.5 GB memory, 2 vCPU burst, one node) without a payment method. Creating anything beyond the free tier needs a payment method on the account, or enough prepaid credit for a new cluster. Changing a cluster's configuration or cloning it needs a payment method or a positive prepaid credit balance. Upgrading the Typesense server version is exempt, and free-tier clusters can always take it. The agent tells you when a payment method is needed and can offer to test with a subset of your data that fits the free tier.
Each connection gets 300 data calls (search, documents, reads and writes) per minute. Cluster-management actions are limited to 30 per minute, the same as the Cloud Management API.
An agent on a server, in CI, or anywhere it cannot open a browser can use a Cluster Management API key as the bearer token on the same URL:
shell
claude mcp add --transport http typesense-cloud https://cloud.typesense.org/mcp/v1 \
--header "Authorization: Bearer ${TYPESENSE_CLOUD_MANAGEMENT_API_KEY}"
Use the key exactly as it was shown when you created it. A key gives access to the cluster tools (whoami, search_typesense_docs, get_clusters, show_cluster and manage_cluster) with the capabilities the key carries. The data tools act as a signed-in person, so they need the browser connection. A headless agent provisions the cluster and issues its keys over MCP, then works with the data through the Typesense Server API directly.
manage_cluster can issue a cluster's admin key and search-only key, both covering every collection and neither expiring. The keys do not appear in the conversation: the agent receives a curl command that downloads them once into a private file on your machine, and reads a key from that file when it needs one. An assistant with no shell, such as claude.ai or ChatGPT, can ask for the keys inline in the response instead.
Keep the admin key on your server and give applications the search-only key. A connection with data:search can also create a search-only key for an application directly on the cluster, if your role manages keys, so the Build preset covers "make my catalog searchable from my app" end to end without issuing an admin key.
For a key limited to certain collections, actions or an expiry, you want to ask the agent to create one. With cluster_credentials:manage, manage_cluster creates a key with exactly the actions and collections you name, with an optional expiry. It hands over the key's value through the same one-time download, or inline for an assistant with no shell. If different users must see different parts of one collection, use a scoped search key.
With the same permission, get_clusters lists the keys on a cluster with their id, description, scope and the first characters of their value. manage_cluster can then revoke a key by its id, including the admin and search-only keys it issued earlier.