Skip to main content
Glama

Plans and issues live on your server instead of in the context window, so work outlives the session. Agents use it over MCP; you use the web UI.

  • 32 MCP tools in 7,290 tokens, about one long file read.

  • One 25 to 32 MB binary with SQLite, the web UI and backups built in. No Docker, no Postgres.

  • lific connect writes MCP config for 11 AI clients.

  • Identifiers like APP-42, never UUIDs.

60-second setup

cargo install lific   # or Homebrew / winget (below), or download a binary from Releases
lific init            # config, database, admin account, background service on :3456
lific connect         # MCP config for your AI clients

Restart your AI client. The web UI is at http://localhost:3456. lific doctor checks config, database, server and an MCP round trip, and exits nonzero if anything is broken.

  • brew install voidnullable/tap/lific replaces cargo install on macOS and Linux: a release binary in seconds, no Rust toolchain.

  • winget install VoidNullable.lific does the same on Windows.

  • lific init --here keeps config and database in the current directory.

  • lific init --no-service skips the background service; run lific start yourself.

  • lific service status | restart | stop | uninstall manages the service.

More: Installation · Quickstart

Related MCP server: MonkeyPlanner

What your agent can do

Call

What it does

get_briefing(project="APP", since="last")

Resume a project: the human queue across projects, what changed since your last briefing, plans and next steps, blocked and workable issues

list_issues(project="APP", workable=true)

Open issues whose blockers are all resolved

list_issues(members=["me"])

Your active work across the projects you belong to

create_plan / get_plan

Nested step plans that survive the session; a step and its issue close together

link_issues(target="APP-9", relation_type="blocks", user="ada")

Block an issue on a person or a date window, not only another issue

update_issue(assignees=["human"])

Mark work a person must do; agents take unassigned issues

update_issue(status="done", evidence="cargo test: 412 passed")

Close with the proof saved as a verification comment

edit_issue / edit_page

Change one string without resending the document

get_activity(identifier="APP", since="2026-10-01")

Who changed what while you were away

Each AI tool connects as its own bot under your account. The history says "OpenCode closed APP-42", and you can revoke one tool without touching the others.

MCP tools

Family

Tools

Issues

list_issues · get_issue · create_issue · update_issue · move_issue · bulk_update · edit_issue · get_board

Relations and waits

link_issues · unlink_issues

Pages

get_page · create_page · update_page · edit_page

Plans

create_plan · get_plan · edit_plan_step · update_plan_step

Comments

add_comment · list_comments · edit_comment · delete_comment

Attachments

upload_attachment · get_attachment · list_attachments

Search and history

get_briefing · search · get_activity

Structure

list_resources · manage_resource · delete

Export

export

The 7,290 tokens are the tools/list definitions as compact JSON, counted with tiktoken's o200k_base. Parameters and output: MCP tools.

Connect your AI tools

lific connect supports opencode · claude-code · claude-desktop · cursor · vscode · codex · zed · gemini · windsurf · goose · crush.

lific connect                                          # pick from the clients it finds
lific connect --client opencode --client cursor --yes  # no prompts
lific connect --client claude-code --scope project     # repo-local .mcp.json
lific connect --client zed --stdio                     # no server: direct SQLite over stdio

It merges into existing config, and leaves a file it can't parse alone and prints the snippet instead. OAuth sign-in from the client and lific login for headless machines also work: Connect agents.

Remote, over Streamable HTTP:

{
  "lific": {
    "type": "remote",
    "url": "http://localhost:3456/mcp",
    "headers": { "Authorization": "Bearer your-api-key" }
  }
}

Local, over stdio with no server:

{
  "lific": {
    "type": "local",
    "command": ["lific", "--db", "path/to/lific.db", "mcp"]
  }
}

Create a key with lific key create --name my-key --user <you>. A stdio session without LIFIC_TOKEN acts as the instance operator.

CLI

lific issue list --project APP
lific issue create --project APP --title "Fix login bug" --priority high
lific issue update APP-42 --status done
lific search "authentication" --project APP
lific --backend http --url https://lific.example.com issue list --project APP   # after lific login

The CLI reads the database directly, or a remote server with --backend http. Output is JSON when piped. lific agents-md adds a short Lific section to a repo's AGENTS.md. Full reference: CLI.

When Lific is the wrong tool

  • You coordinate a large team and need SSO, sprints or Gantt charts. Use Linear or Plane.

  • You want issues stored in the repo itself. Use a git-native tracker such as beads.

  • You need several servers syncing with each other. One Lific instance is one SQLite file.

It's built for one person directing several agents.

Running it

Access. New installs check project membership on every call, reads included. Add people with lific member add --project APP --user sam --role maintainer. See Self-hosting.

Backups. Lific writes an archive every hour and keeps 24. For a one-off:

lific dump                                    # database plus attachments, safe while running
lific restore lific_20260703_141500.tar.gz    # stop the server first

Config. lific init writes lific.toml. Set server.public_url before exposing the server beyond localhost. See Configuration.

Docker. Optional; the binary needs nothing else.

docker build -t lific .
docker run -p 3456:3456 -v lific-data:/data \
  -e LIFIC_INIT_ADMIN_NAME="Your Name" \
  -e LIFIC_INIT_ADMIN_PASSWORD="a long password" \
  lific

Both variables are required on first boot and ignored after it.

Upgrading. Read Upgrading before you skip versions.

Build from source

git clone https://github.com/VoidNullable/lific && cd lific
devenv allow && devenv shell
devenv tasks run lific:debug-build

CONTRIBUTING.md covers tests and release builds.

Community

Questions, bug reports and release announcements: Discord. Support questions get answered fastest in #support.

Contributing

Issues and PRs welcome. For anything large, open an issue first.

MCP Registry

Lific publishes server.json to the MCP Registry.

  • mcp-name: io.github.VoidNullable/lific

License

Apache-2.0

Available Tools

27 tools
add_commentA

Add a comment to an issue (LIF-42) or page (LIF-DOC-3; DOC-3 for workspace pages). The author is the authenticated user.

Inputs: content (required), identifier (required).

bulk_updateA

Apply field changes to matching issues in one call. At most 500 matching issues are selected; narrow filters when more matches exist. Returns the number of issues updated.

Inputs: project (required), set_module, set_status, filter_label, +4 more.

create_issueC

Create a new issue in a project

Inputs: title (required), project (required), labels, module, +5 more.

create_pageC

Create a new page in a project

Inputs: title (required), folder, labels, status, +2 more.

create_planA

Create a nestable step-by-step plan that survives outside the context window. Steps can mirror issues via 'issue': closing the issue completes the step and vice versa.

Inputs: title (required), project (required), steps, anchor_issue.

deleteA

Delete any resource by type and identifier. Types: issue, page, plan, project, module, label, folder.

Inputs: identifier (required), resource_type (required), project.

delete_commentA

Delete a comment by id. Author or admin only.

Inputs: comment_id (required).

edit_commentA

Edit a comment's content by id. Author or admin only; @mentions re-resolve.

Inputs: content (required), comment_id (required).

edit_issueA

Edit an issue by replacing an exact string. Targets the description by default; pass field='title' for the title. Fails if old_string is missing or ambiguous (unless replace_all=true). Cheaper than update_issue for small changes.

Inputs: identifier (required), new_string (required), old_string (required), field, +1 more.

edit_pageA

Edit a page by exact string replacement; same contract as edit_issue. Targets the content by default; pass field='title' for the title.

Inputs: identifier (required), new_string (required), old_string (required), field, +1 more.

edit_plan_stepA

Edit a plan step's text by exact string replacement; same contract as edit_issue. Targets description by default; pass field='title'.

Inputs: plan (required), step_id (required), new_string (required), old_string (required), +2 more.

exportA

Export as markdown: an issue (PRO-42), a page (PRO-DOC-3), or a whole project (PRO). Issues and pages return the markdown; projects return the exported file paths.

Inputs: identifier (required).

get_activityA

Read the audit log: who changed what, when, and through which door (web UI, MCP, API, CLI). Takes an issue, page, or project ID; project scope covers the whole feed. Newest-first with old and new values.

Inputs: identifier (required), limit, offset.

get_boardA

Board view of issues grouped by status (default), priority, or module. Done/cancelled are count-only stubs unless include_closed=true.

Inputs: project (required), group_by, include_closed, max_per_column.

get_issueA

Get an issue by ID (e.g. LIF-1): full details plus the last 3 comments by default.

Inputs: identifier (required), include_comments.

get_pageA

Get a page by identifier (e.g. LIF-DOC-1). Returns full content.

Inputs: identifier (required).

get_planA

Rehydrate a plan's full step tree (e.g. LIF-PLAN-3) when resuming work. Step lines show the #id used by edit_plan_step and update_plan_step, done state, and linked issues.

Inputs: plan (required).

list_commentsB

List comments on an issue (LIF-42) or page (LIF-DOC-3; DOC-3 for workspace pages).

Inputs: identifier (required), limit, order, author, +1 more.

list_issuesC

List issues for a project. workable=true gives issues with no blockers, blocked=true for issues with at least one blocker.

Inputs: project (required), label, limit, order, +11 more.

list_resourcesB

List resources by type: project, module, label, folder, page, issue, or plan. Most types need a project identifier.

Inputs: resource_type (required), label, limit, order, +5 more.

manage_resourceA

Create or update a project, module, label, or folder. Create project requires name and identifier; project update requires project=; module/label/folder create requires project and name; module/label/folder update requires project and current_name. Use delete for deletion.

Inputs: action (required), resource_type (required), name, color, +6 more.

update_issueA

Update an existing issue by identifier. Only provided fields are changed.

Inputs: identifier (required), title, labels, module, +5 more.

update_pageB

Update a page by identifier. Only provided fields are changed.

Inputs: identifier (required), title, folder, labels, +3 more.

update_plan_stepA

Mutate a plan or one step. With step_id: step CRUD, done toggling, attach/detach issue (linked issues sync state). Without step_id: update the plan itself; plan status never closes the anchor issue. Returns a delta.

Inputs: plan (required), done, title, delete, +13 more.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    B3.4/5.0

    Scored across 27 tools

    Disambiguation3/5

    Tools like edit_issue and update_issue have overlapping purposes but are differentiated by method (exact string replacement vs general field update). delete_comment overlaps with the generic delete tool, which can also delete comments, causing potential ambiguity. Overall, most tools are distinct, but these redundancies lower clarity.

    Naming Consistency3/5

    The naming follows a verb_noun pattern for most tools, but there are exceptions like 'delete', 'export', and 'search' without a noun. Additionally, the use of 'edit' vs 'update' for similar operations (edit_issue, update_issue) introduces inconsistency, as does 'add_comment' instead of 'create_comment'.

    Tool Count3/5

    With 27 tools, the count is slightly above the typical range for a well-scoped server. The domain spans issues, pages, plans, comments, resources, and auxiliary features, but redundant tools (delete vs delete_comment, edit vs update for each entity) inflate the count unnecessarily. It's manageable but could be streamlined.

    Completeness5/5

    The tool set covers CRUD operations for issues, pages, plans, and comments, plus resource management, linking, bulk updates, activity logs, board views, export, and search. No major gaps are apparent for the stated domain, and all core workflows appear supported.

    Maintenance

    ActivityNo data
    ResponsivenessOutdated data

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      B
      quality
      A
      maintenance
      A Jira-like project tracker MCP server for AI agents. SQLite-backed, per-project scoped, with full hierarchy and activity logging — so LLMs never lose track. No more scattered markdown files. saga-mcp gives your AI assistant a structured database to track projects, epics, tasks, subtasks, notes, and decisions across sessions.
      41
      76 npm
      43
      MIT
    • A
      license
      Not graded
      quality
      C
      maintenance
      A self-hosted backlog tracker with priority scoring and an MCP server, enabling AI agents to autonomously pull, work on, and update tasks via JSON-RPC tools.
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      A Jira-lite MCP server that enables planning-first workflows for AI-assisted development. It manages tickets as markdown files with YAML frontmatter, guiding agents through epics, stories, tasks, and architecture decision records.
      17
      MIT