M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
I

Inkeep Agents

Official
Framework / SDK
Build AI Agents with a No-Code Visual Builder or TypeScript SDK, with full 2-way sync.

This is a framework for building MCP servers or agents, not a server you connect a client to.

How to install ↓ Compare GitHub source repository ↗
Category
Dev Tools #199 of 438
Stars
★ 1.5k Very popular
Runtime
Node.js
Credentials
No credential needed
License
Other / unspecified
Last commit
54FMRS · D

Inkeep Agents is a fair-code, source-available AI agent platform centered on 2-way sync between a no-code visual builder and a TypeScript SDK. It suits teams that want technical and non-technical collaboration, multi-agent architecture, MCP tools, and self-hosted deployment. It uses ELv2 with Inkeep's Supplemental Terms, which restrict competitive uses; the source provides no client configuration example, so it is not for users expecting a ready-made client setup.

Strongest · Maintenance 13/20 Weakest · Reliability 8/20

Reliability
8/20
Security and permissions
11/20
Maintenance
13/20
Documentation
12/20
Setup experience
10/20
Why each score
Reliability 8/20
This static review can only read the README supplied for the repository root; the server manifest was not cached, so the MCP init handshake, tool listing, and whether declared tools match real behavior cannot be verified. The README describes a multi-service platform (agents-api, agents-manage-ui, agents-sdk, agents-cli, agents-ui) depending on Vercel AI SDK and external LLM providers, implying external runtime dependencies and credentials. No tests, CI workflows, or error-handling descriptions appear in the material. Under the calibration rule, reliability cannot exceed 12 without verifiable execution evidence; the happy path is plausible but edge cases and failure paths are undocumented, so 8.
Security and permissions 11/20
The README mentions 'MCP Tools with credential management', so credentials are in scope, yet nothing discloses storage, encryption, scoping, or least-privilege handling. No tool inventory is available, so it is impossible to judge whether irreversible delete/deploy/pay/remote-command defaults exist, nor whether confirmations or data-flow boundaries are defined. There is no sign of malware, credential theft, covert exfiltration, or fabricated identity, so the 0-4 red line is not triggered; however, missing evidence on permissions and confirmation is a deduction, yielding 11.
Maintenance 13/20
The repository is not archived, shows 1419 stars and 39 open issues, and the README indicates Inkeep ownership with community channels and a contribution guide, suggesting active operation. The license is Elastic License 2.0 plus Inkeep Supplemental Terms (source-available, not fully open source), and repository metadata reports NOASSERTION, leaving licensing boundaries unclear for users. No release history, dependency-update policy, or dedicated security-response channel is visible in the static material, so governance and versioning evidence is thin, giving 13.
Documentation 12/20
The README offers a conceptual overview, quick-start link, SDK code sample, and architecture component descriptions, which is decent onboarding material. It lacks install commands, authentication setup, tool parameter documentation, operational limits, cost notes, and troubleshooting sections; it is also ambiguous whether this is an MCP tool server or a broader platform containing server components, which may misalign capability expectations. Layered docs and disclosed limitations are missing, so per the 'usable but with hidden assumptions' anchor, 12.
Setup experience 10/20
The README points to an external 1-minute quick start, but this repository contains no concrete steps, runtime requirements, dependency installation, or mainstream client configuration examples. The platform is multi-service (API, UI, SDK, CLI) and self-hosting requires Vercel or Docker, so manual configuration dominates. With no reviewable install evidence and the calibration cap of 15 for setup, missing evidence lowers the score, giving 10.

Static review · not runListed 2026-09-10

Read the FMRS scoring method →

Fit and risk

What it can accessRuns commands or codeUses the network

Best for

  • Teams where technical and non-technical members collaborate on agents in one platform
  • Teams needing a no-code visual builder and TypeScript SDK with 2-way sync
  • Teams needing a multi-agent architecture
  • Teams wanting to self-host and deploy on their own infrastructure
  • Developers needing MCP tools with credential management and choice of LLM provider

Not for

  • Users who cannot accept a source-available license or competitive-use restrictions (ELv2 and Inkeep's Supplemental Terms)
  • Users wanting a simple, ready-to-use client configuration (the source provides no client config example)
  • Users unwilling to deploy or maintain infrastructure themselves (self-hosting requires own deployment)
  • Users not intending to use TypeScript or the no-code interface

Required permissions

  • Configuration of agents, sub agents, MCP servers, credentials, and projects (via agents-api)
  • Credential management for MCP tools
  • Access to the configured LLM providers
  • Running on your own infrastructure when self-hosted (Vercel or Docker)
  • Observability data emitted via OpenTelemetry when tracing is used

Risks and side effects

  • ELv2 is a source-available license subject to Inkeep's Supplemental Terms, which restrict certain competitive uses
  • MCP tool credentials, if mishandled, could expose sensitive information
  • Agents can perform write operations such as updating CRMs, knowledge bases, and tickets, so mistakes can affect production data
  • Self-hosted deployment means you bear security and operational responsibility
  • LLM provider data handling and costs depend on the chosen provider

Setup

Before you start

Runtime：Node.js

The server path is the repository root. Per the README, start with the docs (https://docs.inkeep.com) or the 1-minute quick start (https://docs.inkeep.com/get-started/quick-start). There are two ways to build: 1) use the visual builder in agents-manage-ui to drag and drop on a canvas; 2) declare agents in code with the TypeScript SDK (@inkeep/agents-sdk), as shown in the README's basicAgent and helloAgent examples. Use the agents-cli utilities inkeep push and inkeep pull to sync TypeScript SDK code with the visual builder. Deployment can use Vercel or Docker. The source does not provide a specific client configuration example.

Check that it works

Follow the 1-minute quick start (docs.inkeep.com/get-started/quick-start); successfully creating and running a basic agent via the visual builder or TypeScript SDK confirms the setup works.

Troubleshooting

  1. Confirm you have followed the README pointers to the docs and the 1-minute quick start
  2. Check that TypeScript SDK code and the visual builder are correctly synced via inkeep push / inkeep pull
  3. Verify agents-api is running and reachable, since both manage-ui and the SDK write to it
  4. Check MCP server and credential configuration
  5. For self-hosted deployments, verify the Vercel or Docker environment configuration
  6. For execution or evaluation issues, inspect the Traces UI and OpenTelemetry traces

Things to try

Once connected, you can ask your AI assistant things like:

  • Use the Inkeep SDK to create a simple agent that replies with hello world
  • Sync my TypeScript-defined agent to the visual builder with inkeep push
  • Build a multi-sub-agent AI assistant for our help center
  • Embed a chat interface component in my web app

Use cases

Real-time AI chat assistants, such as a customer experience agent for help centers, technical docs, or in-app experiences
An internal copilot to assist support, sales, marketing, ops, and other teams
AI workflow automation such as creating and updating knowledge bases, documentation, and blogs
Updating CRMs, triaging helpdesk tickets, and tackling repetitive tasks
Triggering agents via MCP, A2A, and Vercel SDK APIs
Observability via a Traces UI and OpenTelemetry

Overview

Inkeep Agents is a platform for building AI agents and multi-agent AI workflows. It offers two ways to build: a no-code visual builder (drag-and-drop canvas) and a TypeScript Agents SDK (@inkeep/agents-sdk). The two are fully interoperable with full 2-way sync, so technical and non-technical teams can create and manage agents in one platform. The platform is composed of several services and libraries: agents-api (a REST API that handles configuration of agents, sub agents, MCP servers, credentials, and projects, and also exposes agent execution and evaluation; it tracks conversation state and emits OTEL traces), agents-manage-ui (the visual builder web interface), agents-sdk (the TypeScript SDK for declaratively defining agents and custom tools), agents-cli (utilities including inkeep push and inkeep pull, which sync TypeScript SDK code with the visual builder), and agents-ui (a UI component library of chat interfaces). Under the hood it uses the Vercel AI SDK for interfacing with LLM providers, compatible with Vercel's useChat hook. Agents can be triggered via MCP, A2A, and Vercel SDK APIs; observability is provided via a Traces UI and OpenTelemetry; deployment is possible with Vercel or Docker. The framework is licensed under the Elastic License 2.0 (ELv2) subject to Inkeep's Supplemental Terms, a fair-code, source-available license.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 63.3k · Tools 2 Compare with this →
Compare with other MCP servers

Source revision 63a1b3474154 Data synced 2026-10-19 Read the FMRS scoring method