Lific

Lific is a self-hosted issue tracker built specifically for the emerging workflow where AI coding agents handle development tasks alongside human programmers. Traditional issue trackers assume humans will read tickets, understand context, and execute work manually. They present information through web interfaces designed for visual scanning and clicking, expecting a person to interpret status indicators and navigate between views. Lific instead provides native Model Context Protocol support, allowing AI agents to interact with issues, plans, and comments directly through structured tools. When an agent needs to check what work remains, update progress, or close completed items, it queries Lific through MCP rather than parsing human-readable interfaces. The result is persistent work storage that outlives individual coding sessions and survives context window limits that would otherwise force agents to forget earlier work.

The architecture addresses a practical limitation of current AI coding workflows that anyone using agents seriously has encountered. Agents lose context when sessions end, forgetting what they planned and what remains incomplete. A complex refactoring might span multiple conversations, with each new session starting from scratch unless the human carefully re-explains what came before. Lific stores this information on your server rather than in the agent's context window, creating a durable record that any future session can reference. Issues receive stable identifiers like APP-42 that appear consistently across logs, commit messages, searches, and conversations regardless of when they happen. When an agent or human mentions APP-42, everyone understands which specific piece of work they mean without ambiguity or context-dependent interpretation.

Planning hierarchies allow agents to break complex work into nested task trees where completing child steps automatically closes parent issues. A large feature might decompose into subsystems, each with implementation tasks, and completion flows upward automatically as leaves get done. A workable filter returns only unblocked issues ready for immediate development, preventing agents from wasting precious context tokens on tasks that have dependencies or prerequisites that aren't yet satisfied. The system can generate AGENTS.md files containing tracker instructions formatted for repository integration, helping new agents onboard to a project's workflow quickly without lengthy explanation. The entire MCP interface consumes approximately 5.6k tokens, roughly equivalent to reading a single code file, preserving valuable context space for actual development work rather than tracker navigation overhead.

Human developers aren't afterthoughts in this design despite the agent-centric positioning. The web interface includes a kanban board for visual workflow management, an issue list for comprehensive viewing, document pages for longer-form planning and specification, and discussion threads for ongoing conversation. Project-scoped roles let teams define who can do what within specific projects. OAuth 2.1 authentication integrates with existing identity providers. Audit trails track who did what and when for compliance and debugging purposes. Discussion threads accommodate both synchronous human conversation and asynchronous agent updates, recognizing that future development teams will likely include both human and AI participants contributing at different times and cadences.

For agents that lack MCP support or teams using older tools, a CLI provides identical functionality with JSON output. Every operation available through MCP works through the command line with structured responses that scripts and non-MCP agents can parse. This ensures compatibility across the AI tooling landscape as it evolves, rather than betting everything on MCP becoming the universal standard.

Installation takes roughly one minute through cargo. Running lific init creates configuration files, a local SQLite database, and an API key. The lific connect command configures MCP in detected AI clients automatically by finding their configuration files and adding Lific's server entry. A doctor command validates the entire setup and reports any issues. The application runs on localhost:3456 with automatic service registration, requiring no account creation for local use. Self-hosting means your issue data stays on infrastructure you control rather than flowing through third-party services where it might be used for training, analyzed for metrics, or exposed through breaches.

Lific ships under the Apache 2.0 license with no telemetry whatsoever, meaning it's completely free without hidden data collection or usage tracking. Solo developers, small teams, and organizations needing on-premises infrastructure can adopt it without licensing concerns, usage limits, or surprise charges as workloads grow. The combination of agent-native design, human-friendly interfaces, and zero-cost self-hosting makes Lific a practical choice for teams building workflows where AI agents become regular contributors to codebases alongside their human colleagues.