The agent platform for companies
Agents that work in the software your company already runs on
Elaichi connects 350+ applications and turns them into tools an agent can call from Claude, ChatGPT, Cursor, or any MCP client. Each agent gets exactly the access the person it acts for has. Nothing more.
{
"mcpServers": {
"elaichi": {
"url": "https://api.elaichi.ai/mcp"
}
}
}
// One endpoint for the org, clamped by roles, restrictions and pinned params.
The problem
Most AI tooling gives you reach or control, never both
A company adopts an AI assistant. Someone asks it about a customer and it cannot see the CRM. So IT connects the CRM, and now every employee’s assistant can read every record in it, including records that person could never open in the CRM’s own screen. Elaichi refuses that trade. It connects the same systems, then checks every single call against the live permissions of the person the agent is acting for.
How it works
Connect → Compose → Authorize → Govern
01
Connect
Pick from 350+ applications and finish a hosted sign-in. Credentials go straight into an encrypted vault. Nobody on your team ever handles an API key.
02
Compose
Every connection becomes a set of tools an agent can call. Curate them: rename tools, hide the ones you do not want, pin parameters, and share with a person, a team, or the whole company.
03
Authorize
Point Claude, ChatGPT, Cursor, or any MCP client at the one Elaichi endpoint and sign in with OAuth. No per-toolbox URL, no embedded token. Large toolboxes compact automatically so the model keeps its context.
04
Govern
Admins choose what each role can reach: 36 permissions, 8 predefined roles, and as many custom roles as you need. Every call is checked and logged.
Capabilities
Reach and restraint, in the same platform
350+ applications, nothing to author
Every documented method of every connector is already a tool an agent can call. Connect an account and the toolbox is there. Need something we do not carry? Author a connector from a JSON config, or fork a public one.
Exactly the access you have
Every call is checked against the live role of the person the agent is acting for. Lose access to a system in the morning and the agent loses it on the next request.
Share a capability, not a password
A colleague can run one constrained job on your connection without ever seeing it, holding it, or going beyond it.
Pinned parameters
Fix a tool argument to one value and strip it from the schema the model sees. The agent cannot read it and cannot override it.
One tool that does the whole job
Compose steps across several connections into a single tool, executed server-side under the same checks as everything else.
Big toolboxes stay small
Above about 30 tools, an endpoint offers just search and execute, so a large toolbox never floods the model’s context window.
You choose what the agent can even see
Restrictions work on individual tools, not whole applications, at three levels: organization, role, individual. The individual setting wins, and a blocked tool never appears in the agent’s list.
Every action, attributed
An append-only log records who, or which agent on whose behalf, did what, down to the record changed in the third-party system. Forward it to your Datadog.
The identity stack you already run
SAML and OIDC single sign-on with enforced domains, SCIM v2 provisioning, group-to-role mapping, passkeys, and step-up authentication on sensitive actions.
Want the full tour? Explore the product →
Security
Nobody ever handles a credential
Sign-in happens in a hosted flow. Secrets are encrypted at rest with AES-256-GCM and are never returned by the API, to anyone. Tools run server-side, so no person and no model ever holds a key. Nothing is copied, cached, or indexed either: deals, tickets, and pages are read live from the source when an agent asks for them.
FAQ
Frequently asked questions
What is Elaichi?
Elaichi is an agent platform for companies. It connects the software a business already runs on and lets agents work in those systems, without ever exceeding what the person they act for is allowed to do. Instead of each employee minting ad-hoc MCP servers with personal API keys, Elaichi connects your tools once, turns them into curated toolboxes, and gives every user a personal MCP endpoint clamped by roles, restrictions, and audit logging.
How does Elaichi work?
Four steps. Connect: pick from 350+ connectors and finish a hosted connect flow, where credentials are vaulted and encrypted. Compose: every connection instantly becomes a toolbox of MCP tools, or you curate a template toolbox with renamed tools, pinned parameters, and defaults. Mint: create an MCP endpoint on any toolbox and paste the URL into Claude, ChatGPT, Cursor, or any MCP client. Govern: roles decide who can do what, restrictions decide which tools are even visible, and every call lands in an append-only audit log.
How do I give my team MCP servers without sharing API keys?
You share a capability, not a password. A shared toolbox can expose the sharer's connection: recipients call the third-party API through it without ever seeing credentials, and cannot step outside the tools and parameters you fixed. Every call is clamped server-side by the toolbox's tool list, pinned parameters, and restrictions.
Which AI clients work with Elaichi?
Any MCP client that speaks Streamable HTTP. Claude, ChatGPT, Cursor, and Dust connect directly today. Elaichi MCP endpoints speak standard MCP JSON-RPC (initialize, tools/list, tools/call), and the desktop companion app writes Claude Desktop and Cursor config files for you, so nobody has to hand-edit JSON.
What are pinned parameters?
Pinned parameters are tool arguments fixed by the toolbox curator and stripped from the schema the model sees. They are force-merged at execution time, so the AI can never see or override them. You can pin a Google Drive tool to one specific folder, and the agent can only ever query that folder.
What happens when a toolbox has too many tools for the model's context?
Toolboxes exposing more than about 30 tools automatically advertise just two meta-tools, search_tools(query) and execute_tool(name, arguments), so large toolboxes stay out of the model's context window. This behavior is configurable per MCP server.
Does Elaichi store my SaaS credentials?
Nobody ever handles a credential. Third-party credentials are held in a dedicated credential vault, encrypted at rest with AES-256-GCM, and are never returned by the API, not to users, not to admins, not to agents. Session, API, and MCP tokens are stored as keyed hashes and shown only once.
Can I build a connector for an internal API?
Yes. Custom connectors let you author a connector from a JSON config with base URL, auth, resources, and methods, or fork any of the 350+ public connectors, including its documentation. Forks track upstream lineage, so you can review and selectively pull upstream improvements later.
How is Elaichi different from running my own MCP servers?
Self-hosted MCP servers multiply credentials, have no shared governance, and leave no audit trail. Elaichi centralizes connections with vaulted credentials, lets admins choose what each role can reach, enforces that at connect, advertise, and execute time, logs every tool call, and lets admins revoke or rotate any endpoint instantly.
Put your agents to work in the tools you already use
Start a 14-day Gold trial. Connect a product, curate a toolbox, and paste a governed MCP endpoint into Claude, ChatGPT, or Cursor in minutes.