One MCP server.
1,500+ integrations.
Connect your apps once. Mint a single scoped key. Then point Claude, Codex, Cursor or your own runtime at one endpoint and let it reach 25,000+ tools, without handing any agent a provider credential.
- Free tier
- No credit card
- OAuth handled for you
- Self-host available
# 1. connect an app: VectorHub holds the credential
vectorhub connect salesforce --auth oauth
# 2. scope a key to what one agent may touch
vectorhub keys create support-bot \
--connectors salesforce,slack
# 3. hand that one key to every client you run
claude mcp add vectorhub --transport http \
https://mcp.vector-hub.org/v1 \
--header "Authorization: Bearer $VECTORHUB_KEY"
✓ 1,847 tools available to support-bot- 1,500+
- Managed connectors
- 25,000+
- Callable tools
- 1
- Endpoint to configure
- 0
- Credentials in your agent
Connectors maintained by VectorHub
Every new agent restarts the integration work
Three agents that each need twenty tools is not twenty pieces of work. It is sixty, plus sixty credentials, sixty error paths and sixty things that break when an upstream API changes.
The N × M problem
Five agents that each need twenty tools is a hundred integrations to build, secure and maintain. Every new agent starts the work over.
Credentials sprayed everywhere
Provider tokens end up in dotfiles, CI variables and three different laptops. Rotating one means hunting down every copy.
Context windows blown on schemas
Loading a few hundred tool definitions costs tens of thousands of tokens before the agent has done anything useful.
Integrations rot silently
APIs change. Hand-rolled clients break on a Friday and nobody notices until an agent quietly starts failing every task.
Your agents
One URL + one scoped key each
VectorHub Gateway
mcp.vector-hub.org/v1
Scope check
Credential injection
Tool discovery
Audit log
Your systems
Real credentials never leave the vault
Four steps, then never again
Connect apps once at the gateway. Every agent you add after that costs one command.
Connect your apps once
Authenticate each service in the VectorHub dashboard: OAuth consent screen, API key, username and password, or a custom header scheme. Credentials are encrypted and stored server-side.
# or do it from the CLI
vectorhub connect salesforce --auth oauth
vectorhub connect stripe --auth api-key
vectorhub connect internal-billing --auth basicMint a scoped key
Create a key bound to exactly the connectors and tools one agent should reach. A support bot gets Zendesk read plus Slack write, and nothing else: least privilege by construction.
vectorhub keys create support-bot \
--connectors zendesk,slack \
--tools "zendesk.ticket.*,slack.message.post" \
--scope read,write
vh_live_9f2c… # copy once, rotate any timeAdd one MCP server to every client
Claude, Codex, Cursor, Windsurf, LangChain, your own runtime: each gets the same URL and its own key. No per-agent OAuth, no secrets in dotfiles.
claude mcp add vectorhub \
--transport http \
https://mcp.vector-hub.org/v1 \
--header "Authorization: Bearer $VECTORHUB_KEY"Watch it work
Every tool call is traced with arguments, latency and result status. Revoke a key and access stops everywhere it was used, instantly.
12:04:31 salesforce.opportunity.search 187ms ok
12:04:32 slack.message.post 92ms ok
12:04:36 stripe.invoice.create 241ms okEverything between your agent and your systems
The gateway is not a proxy with a nice logo. It is the auth layer, the discovery layer and the audit layer that agent tooling has been missing.
One endpoint, every integration
Point an agent at a single MCP URL. Behind it sit every connector you have enabled: add Salesforce on Tuesday and the agent can call it on Tuesday, with no client-side change.
Learn more1,500+ connectors, maintained for you
Schemas track upstream APIs. When a vendor deprecates a field or rotates a version, the connector is updated centrally instead of breaking in your codebase.
Learn moreTurn any API into an MCP server
Paste an OpenAPI or GraphQL schema and get a typed MCP server with tools, descriptions and auth wired in. Internal services included.
Learn moreCredentials live in the vault, not the agent
OAuth 2.0, API keys, basic auth and custom headers are stored encrypted and injected at call time. Your agent never receives a provider secret.
Learn moreSemantic tool discovery
25,000 tools will not fit in a context window. VectorHub ranks and exposes only the tools relevant to the task, so prompts stay short and cheap.
Learn moreEvery call logged and replayable
Structured logs, latency traces and per-tool error rates for every invocation, filterable by key, connector and agent.
Learn moreFour ways in. One way out.
OAuth, API keys, username and password, or something bespoke: VectorHub speaks all of them upstream, and exposes exactly one bearer token downstream.
OAuth 2.0
Full authorization-code flow with PKCE, refresh-token rotation and per-user consent. VectorHub handles the callback, the token store and the refresh loop.
API keys & tokens
Static keys, bearer tokens and signed headers, stored encrypted and injected at call time. Rotate from the dashboard without redeploying anything.
Username & password
Basic auth and form-login services, including legacy internal systems that never got a modern auth story.
Custom schemes
HMAC request signing, mTLS client certificates and multi-step token exchanges, described declaratively per connector.
1,500+ integrations across 12 categories
Every connector ships typed tools, versioned schemas and an auth binding. When an upstream API changes, the fix lands centrally instead of in your codebase.
Productivity
18+Docs, tasks, calendars and notes.
CRM & Sales
12+Pipelines, contacts and revenue data.
Developer Tools
20+Repos, CI, incidents and deploys.
Communication
14+Chat, mail, meetings and calls.
Data & Analytics
16+Warehouses, BI and event streams.
Marketing
12+Campaigns, ads, SEO and content.
Finance & Billing
11+Payments, ledgers and invoicing.
Support
7+Tickets, help centers and CSAT.
Files & Storage
7+Object stores and file sync.
HR & Recruiting
8+ATS, HRIS and onboarding.
Commerce
7+Storefronts, orders and inventory.
AI & ML
9+Models, vector stores and evals.
Add it once to every client you use
VectorHub speaks MCP over streamable HTTP, so any compliant client works. Here is the exact configuration for the ones teams use most.
claude mcp add vectorhub \
--transport http \
https://mcp.vector-hub.org/v1 \
--header "Authorization: Bearer $VECTORHUB_KEY"Using something else? See all client integrations.
What teams build on one key
Cross-system work is where agents earn their keep, and exactly where per-service integrations fall apart.
Revenue operations agent
Reads pipeline from Salesforce, enriches from Apollo, posts a digest to Slack and files follow-ups in Linear: one key, four connectors.
Support triage agent
Classifies incoming Zendesk tickets, pulls order state from Shopify, refunds through Stripe when policy allows and escalates the rest.
Engineering on-call copilot
Correlates a PagerDuty incident with Sentry errors and Datadog metrics, finds the offending deploy in GitHub and drafts the postmortem.
Analytics answer bot
Answers questions in Slack by querying Snowflake, cross-checking the dbt model and linking the Looker dashboard it came from.
Rolling your own vs. the gateway
Both paths work on day one. They diverge sharply on day ninety.
Frequently asked questions
The short version of what teams ask us before they connect their first app.
What is an MCP server, and why does one gateway help?
The Model Context Protocol is an open standard for exposing tools to AI agents. Normally each service needs its own MCP server, its own credentials and its own client configuration. VectorHub runs one server that fronts every connector you enable, so an agent configures a single endpoint and gains access to everything behind it.
Do I need to give every agent its own credentials?
No. That is the point of the gateway. You authenticate each service once in VectorHub, then mint a scoped key per agent. Claude, Codex, Cursor and your own runtimes each get a key that grants only the connectors and tools you chose. Provider secrets never leave the vault.
How do 25,000 tools fit in a context window?
They do not, which is why VectorHub does not send them all. Tool discovery ranks the catalog against the current task and exposes a working set, so the agent sees the tools it needs rather than every tool that exists.
Can I connect an internal API that has no public connector?
Yes. Point the API → MCP converter at an OpenAPI or GraphQL schema, or describe the endpoints manually, and VectorHub generates a typed MCP server with auth attached. Private connectors stay scoped to your organisation.
What happens when I revoke a key?
Access stops immediately, everywhere that key was configured. Because agents hold a VectorHub key rather than provider credentials, revocation is a single action instead of a rotation project across every service.
Which authentication methods are supported?
OAuth 2.0 with PKCE and refresh rotation, API keys and bearer tokens, username and password for basic-auth services, and custom schemes such as HMAC request signing or mTLS.
Can VectorHub run inside our own infrastructure?
Yes. Enterprise deployments can run in your VPC or fully self-hosted, with data residency controls and audit export. Talk to us about the deployment model you need.
Is there an SDK, or is it MCP only?
MCP over streamable HTTP is the primary interface and works with any compliant client. A REST API and TypeScript and Python SDKs cover provisioning, key management and log access.
Give every agent one endpoint
Connect your apps once, mint a scoped key, and point Claude, Codex, Cursor or your own runtime at a single MCP server.
No credit card required · Free tier available · Self-host on request