Kaiku

Product Introduction

1. Definition: Kaiku is an agent-native, API-compatible project management and wiki platform. Technically, it is a SaaS-based task tracker and knowledge base built with a first-class Model Context Protocol (MCP) integration layer, designed to function as a drop-in replacement for incumbent enterprise tools like Jira and Confluence.
2. Core Value Proposition: Kaiku exists to seamlessly integrate AI agents into human-centric workflows. Its primary value is providing a unified workspace where AI agents can operate with full transparency and human oversight, using the same APIs and protocols as existing enterprise tools, thereby eliminating integration lag and vendor lock-in for AI-augmented teams.

Main Features

1. Native MCP Integration & Wire-Level Compatibility: Kaiku speaks the Model Context Protocol (MCP) natively in two ways. First, it provides its own built-in MCP server, allowing agents direct access to issues, comments, wiki pages, and attachments. Second, and crucially, it offers wire-level compatibility with MCP servers built for other major trackers (e.g., Jira) and wikis. This means agents configured for those platforms can connect to Kaiku's REST API endpoint without modification, as the request/response shapes and error formats are identical. This eliminates the need for lagging bridge software.
2. Transparent, Cost-Attributed Agent Operations: Every AI agent interaction is logged as a discrete "run" directly on the associated issue or page. Each run record includes granular cost breakdowns: tokens consumed (by type, e.g., input/output), compute time, and the identity of the human user who initiated the call. This provides real-time visibility into the operational costs of automation, enabling proactive budget and workflow management.
3. Proposal-Only Agent Framework with Human-in-the-Loop: Agents in Kaiku are roles, not accounts, and execute with the permissions of the user who calls them. When invoked via a comment (e.g., @agent), the agent reads the issue context using its tools (which are read-only) and posts a proposal as a comment. This proposal is recorded as a formal question addressed to a human, requiring explicit approval. The system enforces a "propose, never decide" paradigm, ensuring human accountability and preserving an audit trail of who approved what action.
4. Unified Wiki with Diagram OCR and Standardized Export: Each project includes a dedicated wiki space. Beyond standard Markdown editing, Kaiku includes advanced features like Optical Character Recognition (OCR) on attached diagrams, making text within images searchable. Wiki pages export to Word (.docx) format, and the platform's storage layer uses the same wire format as major corporate wikis, ensuring compatibility with third-party tools built for those ecosystems.
5. Mini-CRM Module with Custom Fields: Beyond standard backlog management, projects can enable a "client book" view. This transforms the project into a lightweight CRM where each client is a specialized issue type. Teams can define custom column fields (e.g., colored lists, dates, numbers, countries) with strict validation. The entire client book can be exported to .xlsx format with preserved data types, and edits can be made directly in-table, syncing instantly with the underlying issue data.
6. Data Sovereignty and Graceful Degradation: Kaiku emphasizes user data ownership. A single API call exports an entire project as a ZIP archive containing structured JSON, Markdown files, and all attachments. If a subscription lapses, the workspace becomes read-only; data is never deleted or held hostage. The platform maintains full functionality for 12 months in this state, allowing for data retrieval.

Problems Solved

1. Pain Point: Disconnected AI tooling and opaque automation costs. Teams using AI agents face a fragmented experience where agents operate in silos, using separate MCP servers or custom integrations. The cost and activity of these agents are not tracked within the core project management tool, leading to hidden expenses and unclear ROI.
2. Target Audience: Development teams, engineering managers, and product teams who are increasingly incorporating AI agents (e.g., using Claude Code, Cursor, or custom agents) into their development and project management lifecycle. It also appeals to companies frustrated with the slow pace of AI integration in legacy enterprise SaaS platforms.
3. Use Cases: A software team can connect their existing Claude-for-Jira MCP server directly to Kaiku; their agents continue working uninterrupted while all activity and costs are logged per issue. A product manager can @mention an agent in a comment to draft a product requirements document in the wiki based on linked issues, review the proposal, and approve it—all within a single thread. A small business can use the Mini-CRM to manage client projects, using custom fields to track deal size and status, with all client communication and tasks nested underneath.

Unique Advantages

1. Differentiation: Unlike traditional project management tools that treat AI as a bolt-on feature or chatbot, Kaiku is architected from the ground up for agent interaction. Unlike building custom integrations, it offers immediate, zero-config compatibility with the existing ecosystem of MCP servers. Compared to other "AI-native" tools, its rigorous human-in-the-loop approval and detailed cost attribution are core, non-optional features.
2. Key Innovation: The dual-mode MCP compatibility is its key technical innovation. By implementing the wire protocol of incumbent tools, Kaiku achieves backward compatibility without maintaining fragile, version-lagged API bridges. This allows teams to adopt an agent-native platform without disrupting their existing automated workflows, a significant barrier to entry for new tools.

Frequently Asked Questions (FAQ)

1. Is Kaiku compatible with my existing AI agent setups for Jira or Confluence? Yes, Kaiku offers direct wire-level compatibility. If your AI agent (e.g., in Claude Desktop or Cursor) uses an MCP server configured for Jira Cloud's REST API, you can simply point it to your Kaiku workspace URL and token. The agent will interact with Kaiku as if it were Jira, with no code changes required.
2. How does Kaiku prevent AI agents from making unauthorized changes? Kaiku enforces a strict "propose, never decide" model. Agents have no accounts or direct write access. They execute with the permissions of the user who calls them, and their tools are read-only. Any action an agent suggests is posted as a proposal comment that must be explicitly approved by a human user, creating a clear audit trail.
3. What happens to my data if I cancel my Kaiku subscription? Your data is never locked. Upon subscription lapse, your workspace becomes read-only for 12 months. All data remains accessible and exportable during this period. You can perform a full project export (including structured JSON and files) at any time to migrate your data.
4. Can I use my own AI API keys with Kaiku? Yes. The "Team, own AI key" plan allows you to bring your own Anthropic or DeepSeek API keys. All agent runs, the AI assistant, and automatic translations will use your key, and you will be billed directly by the AI provider. Kaiku charges a reduced platform fee in this model.
5. How does the Mini-CRM differ from a full CRM like Salesforce? Kaiku's Mini-CRM is a project management feature, not a standalone sales platform. It is a structured table view of client records (which are issues) within a project, with custom fields and nested tasks. It lacks dedicated sales pipeline automation, email sequencing, or call logging. It is designed for teams who need to track client-related work within their primary task tracker, not replace a specialized sales CRM.