Knownbase is a persistent, searchable project memory for AI coding agents. It helps Claude Code, Codex, Cursor, ChatGPT, and other MCP-compatible agents retain important project knowledge across sessions instead of repeatedly rediscovering the same context.AI coding agents are effective within a session, but valuable knowledge often disappears when the session ends, context is compacted, or work moves to another agent. Architecture decisions, debugging discoveries, rejected approaches, deployment constraints, conventions, and implementation details can end up buried in chat histories or scattered across oversized instruction files.Knownbase gives that knowledge a durable place to live.Key FeaturesPersistent Project Memory: Preserve important knowledge beyond individual conversations, sessions, and context windows.Cross-Agent Memory: Knowledge captured while working with one agent can be retrieved by another MCP-compatible agent.On-Demand Retrieval: Agents search for relevant project knowledge when needed instead of loading an ever-growing memory file into every prompt.Structured Knowledge: Store architecture decisions, debugging trails, operational constraints, conventions, handoffs, rejected approaches, and other project facts.Searchable & Organized: Organize knowledge by project with statuses, tags, Markdown content, linking, and revision history.MCP Integration: Connect compatible AI tools through a standard Model Context Protocol endpoint.Human + Agent Workspace: Developers and AI agents can contribute to and retrieve from the same project knowledge base.How Knownbase HelpsConsider a long debugging session where an agent discovers that a production issue is caused by a specific interaction between a database migration, cache behavior, and an older compatibility requirement.That discovery may be critical months later.Without persistent memory, the next agent may inspect the same code, reproduce the same investigation, and potentially suggest an approach that was already tested and rejected.With Knownbase, the useful discovery can be stored as durable project knowledge. A future session can search the project memory, retrieve the previous findings, and continue from there.This is particularly useful for:Long-running AI-assisted developmentContext compaction and session resetsArchitecture and technical decisionsDebugging discoveries and incident historyAgent-to-agent handoffsProject conventions and constraintsRecording rejected approaches and why they failedSwitching between Claude Code, Codex, Cursor, ChatGPT, and other MCP clientsSharing accumulated project knowledge across a development teamShared Memory Across AI ToolsKnownbase is not tied to the lifetime of one chat or one AI coding tool.A discovery stored during a Claude Code session can later become useful context for Codex, Cursor, ChatGPT, another MCP-compatible agent, or a teammate working on the same project.This creates a persistent knowledge layer between your codebase and the agents working on it:Agent discovers → Knownbase stores → Future agent searches → Relevant context is retrieved → Work continuesInstead of every agent starting from zero, useful project knowledge accumulates over time.Designed for Selective ContextPersistent memory should not mean injecting the entire history of a project into every prompt.Knownbase keeps project knowledge outside the agent's immediate context window. Agents can search and retrieve relevant information when they need it, keeping working context focused while maintaining access to a much larger body of historical knowledge.Simple MCP IntegrationKnownbase is delivered as a hosted MCP service. Connect your compatible AI client to the Knownbase MCP endpoint, authenticate it, and the agent can use Knownbase's available memory tools to search, read, create, and maintain project knowledge.There is no SDK or vector database infrastructure for developers to operate themselves, and Knownbase does not require modifying the repository simply to provide persistent memory.PricingKnownbase includes a Free plan with 3 projects, 300 notes, and 2 MCP keys, with no credit card required.Paid plans include a 7-day free trial, allowing developers and teams to evaluate larger-scale workflows before committing.Current LimitationsKnownbase is focused specifically on durable, searchable project memory rather than trying to replace every collaboration or documentation tool.Current limitations include:AI agents must support MCP and be configured to use Knownbase.Memory quality depends on useful information being stored during the development workflow.Knownbase is currently a hosted service rather than a self-hostable open-source server.It is not intended to replace Git, repository documentation, issue trackers, or real-time collaborative editors.Instead, Knownbase complements those systems by preserving the working knowledge agents repeatedly need but conventional source control often does not capture.Why Knownbase?Your repository stores the code.Git stores how the code changed.Documentation describes how the system is supposed to work.Knownbase preserves what your AI agents learn while working on it.That includes the decisions, discoveries, constraints, failed approaches, debugging history, and project-specific knowledge that would otherwise disappear into old conversations.Stop making every AI coding session rediscover your project.Give your agents a persistent, shared memory they can search when they need it.