One MCP endpoint / 01
Connect once.Swap agents freely
SPYN3 puts one MCP endpoint in front of every system you connect — the MCP servers you already use, plus Postgres, REST APIs and Google Drive. Point Claude, Cursor or whatever ships next at one URL, and take it with you when you switch.
Move one URL, not every integration.
Free plan · 500 tool calls a month · no card
60 seconds
One endpoint, end to end
Real screens from the dashboard and a real agent session through SPYN3: connect sources, point an agent at one URL, switch agents without touching an integration.
The problem
Every new agent, editor, or automation makes you re-connect and re-authorize the same systems. The result is repeated integration work, credentials scattered across tools, a tool list that grows until the agent chooses badly, and no shared permission model.
How SPYN3 works
Three steps to an agent you can replace
Connect
Add the MCP servers you already run, plus the systems that don't speak MCP — a Postgres database, any REST API. Encrypted, read-only, permission-scoped.
SPYN3 federates
Every source's tools appear through one endpoint, foldered and namespaced, so your agent sees a short list and opens what it needs.
Point any agent at it
Claude, Cursor, whatever ships next quarter — one URL. Switching agents means moving one line, not rebuilding every integration.
What you get
Infrastructure, not another assistant
SPYN3 is the layer underneath your agents — the integrations you'd otherwise wire into each one separately, and rewire on every switch.
One endpoint, every system
A hosted MCP server per workspace, in front of everything you've connected. Unplug it from one agent and plug it into another.
A tool list that stays short
Five servers with twenty tools each is a hundred tools, and agents choke on that. Tools are foldered by source and revealed only when opened.
One permission model
SPYN3 is designed around workspace isolation, hashed API keys, and encrypted connector credentials. Scope each agent to only the access it needs.
Token-shaped returns
Agent-facing responses serialize minimal-by-default with hard caps — so you spend context on thinking, not on plumbing.
Where this goes / 04
Tools and memory.One endpoint.
Memories, skills and notes live behind the same endpoint as your tools, so Claude Code, Codex, Cursor and whatever comes next start from the same context. Off until you turn it on; your personal memory stays yours. How memory works
For builders
Your agent, one connection away
Register SPYN3 as an MCP server once. Your agent sees a short list — each connected source is a folder — opens the one it needs, and calls its tools under a namespace, so two upstreams can both have a search.
Support the build
Back an independent developer
SPYN3 is built solo, in the open. If the vision resonates, you can chip in to fund the infrastructure and buy some late-night tea — no account needed, pay what you like.
Support SPYN3 ↗Pricing
Pay for what your agents call
Pro
$29 / month
- 10,000 tool calls included
- 3 workspaces, unlimited sources
- 10 API keys
- $0.002 per call beyond that
Team
$99 / month
- 50,000 tool calls included
- Unlimited workspaces, sources and keys
- 5 seats
- $0.002 per call beyond that
A tool call is one call your agent makes through SPYN3 to a connected source. Browsing folders is free. Upgrade from the dashboard — or ask your agent to, with upgrade_plan.
Questions
What people ask about SPYN3
What is SPYN3?
SPYN3 is a consolidated MCP layer: one endpoint in front of every system you connect. Your agent talks to SPYN3, and SPYN3 talks to GitHub, Linear, Notion, your Postgres database and the rest. When you switch agents, you move one URL and every integration comes with it.
How is SPYN3 different from adding MCP servers to my agent directly?
Connected directly, every agent holds its own copy of every connection and credential, and its tool list grows with each server — five servers with twenty tools each is a hundred tools, and agents choose worse on long lists. SPYN3 holds each connection once, folders tools by source and reveals them only when opened, and puts one permission model and one audit log over all of them.
Which agents work with SPYN3?
Any MCP client that can connect to a remote (Streamable HTTP) server — Claude, Cursor and others. Agents authenticate with a workspace API key or with OAuth.
What can I connect to SPYN3?
Remote MCP servers: a gallery of 44 well-known ones — including Notion, Linear, Atlassian, Sentry, Supabase, Stripe and GitHub — most of which sign in with OAuth, or any other server by URL. Systems that don't speak MCP — a Postgres database, any REST API, Google Drive — are exposed as read-only tools.
How does SPYN3 keep my systems and data safe?
Workspaces are isolated with Postgres row-level security. API keys are stored as hashes, connector credentials are encrypted at rest, and each key can be scoped to only the sources it needs. Postgres runs inside read-only transactions and REST sources are GET-only, so an agent cannot write through them. Every proxied call is logged with its source and latency. SPYN3 passes calls through; it does not store or index the content.
How much does SPYN3 cost?
Free covers 500 tool calls a month with no card. Pro is $29 a month with 10,000 calls included; Team is $99 a month with 50,000 calls and five seats. Beyond the included calls, paid plans cost $0.002 per call.
Start free
Stop rewiring your agents
Connect a source, point your agent at one URL, and see whether it earns a place in your setup. 500 tool calls a month are free.