project-progress
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@project-progresslist all my projects and their status"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Resumable memory for Codex, Claude Code, and any MCP-compatible agent.
Project-local Markdown that survives context resets — so your agent picks up exactly where it left off, every session, in one file it can read in a few hundred tokens.
Why · Features · Quick Start · MCP Tools · Troubleshooting
Why
Agent sessions end. Context windows reset. New sessions start cold, burning the first few minutes re-deriving "what was I doing?"
Awesome Progress Tracker fixes that with one convention: every project keeps a project-progress/Progress.md file as its resume source of truth. A SessionStart hook injects the compact bits — Resume Snapshot, Next Action, Blockers — into the agent's context automatically. No dashboards, no databases. Just a Markdown file the agent reads at kickoff and updates before it stops.
session 1 ──work──► Progress.md updated (Resume Snapshot · Next Action · Blockers)
│
new session ◄── SessionStart hook injects the snapshot, no lookup needed ──┘Related MCP server: agent-coordination-mcp
Features
🗂️ One file per project, human-readable — plain Markdown you can read, edit, diff, and commit.
🔁 Automatic resume context —
SessionStartinjects the Resume Snapshot and Next Action at kickoff.🛡️ Sensitive-commit guard — blocks
git commitwhen staged progress is markedcommit_progress: falseorsensitivity: sensitive.⏰ Stop reminders — flags when the working tree changed but
Progress.mddidn't, and scans for accidentally-committed secrets.🧩 Works everywhere — first-class plugins for Claude Code and Codex, plus an open-standard
SKILL.mdskill for Gemini CLI, Copilot, and Cursor.🌐 Cross-project index via MCP — answers "what am I working on?" across every tracked project (see MCP Server).
🩺 Scriptable health checks —
doctor --jsongives a non-zero exit code and a diagnostic report when setup is broken.
Choose Your Path
I use... | Install this way |
Claude Code | Claude Code plugin — zero-config, one marketplace add |
Codex | Codex plugin — native plugin flow, no |
Hermes Agent | Hermes install — Skill + MCP only; lifecycle hooks are deferred |
Gemini CLI / Copilot / Cursor | Drop |
MCP client or CLI only | Manual install and, if needed, MCP Server |
Requires Node.js >= 18. Hooks and MCP server run entirely on Node — no Python needed.
After installing a plugin, start a new agent session. Run awesome-progress-tracker doctor -g codex (or --json for scripting) to check setup.
Install As A Claude Code Plugin (recommended)
Zero-config: one marketplace add + install wires the skill, MCP server, and lifecycle hooks — no edits to CLAUDE.md or ~/.claude.json.
/plugin marketplace add AndriiLavrekha/awesome-progress-tracker
/plugin install project-progress@awesome-progress-trackerOr from the CLI:
claude plugin marketplace add AndriiLavrekha/awesome-progress-tracker
claude plugin install project-progress@awesome-progress-trackerWhat you get (restart the session to load it):
Piece | What it does |
Skill | Guides the agent to maintain |
Command | Initializes tracking in the current repo |
MCP server |
|
| Injects the Resume Snapshot / Next Action as context |
| Blocks |
| Reminds you to update progress (and flags secrets) when the tree changed but |
To scope the MCP index to specific roots, set PROJECT_PROGRESS_ROOTS (semicolon-separated) in the environment Claude Code runs in.
Ships prebuilt
dist/, so it runs straight from the cloned repo with no build step.
Install As A Codex Plugin (recommended)
The same repo is also a Codex plugin — installs through the native flow, no manual config.toml editing:
codex plugin marketplace add AndriiLavrekha/awesome-progress-tracker
codex plugin add project-progress@awesome-progress-trackerThis wires the same skill, MCP server, and lifecycle hooks as the Claude Code plugin. Restart Codex after installing or updating so the manifest, MCP server, and hooks reload.
Codex doesn't auto-trust plugin-bundled hooks: the first timeSessionStart / PreToolUse / Stop fire, Codex asks you to review and trust them. Approve once.
Codex
codex plugin marketplace upgrade
codex plugin add project-progress@awesome-progress-tracker
codex plugin listStart a new Codex session afterward. If prompted, review and trust the updated hooks in /hooks.
Claude Code
claude plugin update project-progress@awesome-progress-tracker
claude plugin listRestart Claude Code so the plugin's skills, MCP server, and hooks reload.
On an initialized project, SessionStart injects resume context from Progress.md. On an uninitialized project, the agent asks before creating project-progress/:
This project is not initialized with Awesome Progress Tracker. Do you want me to create `project-progress/` here?Yes → run
awesome-progress-tracker init . --project "<name>"No → record an opt-out so future sessions stay quiet:
awesome-progress-tracker state set . --state opted-outThe opt-out lives outside the repo, under the tool's user data directory — it never dirties the project.
Because skills follow the open SKILL.md standard, the same skill also works in Gemini CLI, Copilot, and Cursor when placed under ~/.agents/skills or a repo's .agents/skills.
Install For Hermes Agent
Hermes Agent is supported today through the managed skill + MCP path: supported Skill + MCP installation with the install, doctor, and uninstall commands, plus status for inspection. lifecycle hooks are deferred for Hermes, so this path does not yet wire SessionStart, PreToolUse, or Stop automation.
If Hermes isn't installed yet, follow the official Hermes Agent installation instructions, then confirm the CLI works:
hermes --versionWire the integration:
npx github:AndriiLavrekha/awesome-progress-tracker install -g hermesPiece | What it does |
Skill | Installed from the tagged raw GitHub |
MCP server | Added via |
Doctor | Checks Hermes CLI, skill, MCP presence, and |
Uninstall | Removes only the managed skill and MCP server |
Named collisions stop the install before any changes are made. If Hermes already has a project-progress skill or awesome-progress-tracker MCP server, remove or rename the existing entry first and rerun the installer.
After install or update, restart Hermes so it reloads the managed skill and MCP registry. Because lifecycle hooks are deferred for Hermes, drive things manually for now:
initialize with
awesome-progress-tracker init . --project "<name>"when the user opts incheck setup with
awesome-progress-tracker status -g hermes/doctor -g hermesverify directly with
hermes skills list --source hub,hermes mcp list,hermes mcp test awesome-progress-tracker
Initialize A Project
# after npm publication
npx awesome-progress-tracker init /path/to/repo --project "My Project"
# from the private GitHub repo
npx github:AndriiLavrekha/awesome-progress-tracker init /path/to/repo --project "My Project"
# or installed globally
npm install -g github:AndriiLavrekha/awesome-progress-tracker
project-progress init /path/to/repo --project "My Project"init creates project-progress/ from templates/project-progress/. Fill in the frontmatter and keep Resume Snapshot, Next Action, Remaining Work, and Blockers compact enough for an agent to load first.
Install the global bootstrap for your agent (claude is the default; swap -g codex where shown). This installs global instructions telling the agent to check for project-progress/Progress.md at kickoff — it does not initialize every project automatically. It also configures the agent's MCP client (~/.claude.json for Claude Code, ~/.codex/config.toml for Codex).
Task | Command |
Install |
|
Install + verify |
|
Install MCP config only |
|
Install MCP config, project-local |
|
Scan other roots |
|
Check status |
|
Health check |
|
List opt-in/opt-out state |
|
Set opt-out |
|
Reset state |
|
Uninstall |
|
Replace ... with npx github:AndriiLavrekha/awesome-progress-tracker.
Manual instruction files are also available:
skills/project-progress/SKILL.md— install or reference for Codexagent-instructions/AGENTS-snippet.md— paste into a project or global AGENTS.mdagent-instructions/CLAUDE-snippet.md— paste into Claude Code memoryagent-instructions/HOOKS.md— lifecycle reminders and validation
Agents should update progress at kickoff when state changes, after milestones, when blockers appear, after verification, and before ending a meaningful session.
# Windows
./hooks/project-progress-check.ps1 -ProjectRoot . -SessionStartedAt 2026-06-27T00:00:00+00:00 -MeaningfulWork -CompletionBoundary# POSIX
./hooks/project-progress-check.sh --project-root . --session-started-at 2026-06-27T00:00:00+00:00 --meaningful-work --completion-boundaryThese wrappers run the compiled hook (dist/src/hook/cli.js) on Node — run npm run build first (or install the published package, which builds on prepare). For direct use: node dist/src/hook/cli.js --project-root . --session-started-at <iso>.
MCP Server
npx awesome-progress-tracker mcp # after npm publication
npx github:AndriiLavrekha/awesome-progress-tracker mcp # from the private GitHub repoConfigure discovery with a semicolon-separated PROJECT_PROGRESS_ROOTS:
$env:PROJECT_PROGRESS_ROOTS = "C:/Users/you/Documents;C:/Users/you/Projects"
npx github:AndriiLavrekha/awesome-progress-tracker mcpFor local development in this repo:
npm install
npm run build:mcp
node dist/src/mcp/server.jsMCP clients installed from npm/GitHub should use the package binary (awesome-progress-tracker mcp or project-progress mcp) to keep stdio clean.
Tool surface — deliberately small; administrative state lives in the CLI (state list/set/reset), not here:
Tool | What it does |
| List compact summaries from the cached index; optional |
| Rescan |
| Read one project's compact progress summary |
| Replace or append a named section in a project's |
| Update frontmatter |
The server maintains a lightweight global index (~/.awesome-progress-tracker/projects.json and Projects.md) as a fast "what projects exist?" view. Progress.md remains the source of truth; refresh_projects, init, update_project_progress, and mark_project_status all keep the index in sync.
Verification
See TESTING.md and agent-instructions/SELF-TEST.md for full scenarios.
npm test
npm run buildVerify package creation and npx-style execution locally:
npm pack --dry-run
npm pack
npx --yes ./awesome-progress-tracker-0.1.0.tgz helpRun the lifecycle check against this repo:
./hooks/project-progress-check.ps1 -ProjectRoot . -SessionStartedAt 2026-06-27T00:00:00+00:00 -MeaningfulWork -CompletionBoundaryTroubleshooting Codex
Check in order:
Plugin installed and enabled:
codex plugin list.Hooks trusted: open
/hooksin Codex and trust theproject-progresshooks.Restarted Codex/Claude Code (or new session) after install or update.
Project is actually uninitialized:
project-progress/Progress.mdis missing.Project isn't opted out:
awesome-progress-tracker state list; reset withstate reset ..Task is non-trivial: hooks only nudge for multi-step feature/investigation/refactor/setup/debugging/deployment/release work.
MCP is running:
/mcpin Codex orawesome-progress-tracker doctor -g codex.
The prompt-time hook covers work submitted after startup — if the project was created mid-session, submit the next prompt after trusting the updated hook.
Hooks are best-effort and never block normal Codex operation. If hooks are disabled or untrusted, the project-progress skill and bootstrap instructions still define the workflow.
Contributing
Private, single-maintainer repo for now. See AGENTS.md for structure/conventions and TESTING.md for the verification workflow before opening a PR.
Each project owns its own progress files — global vaults, dashboards, and MCP tools may read or summarize them, but the source of truth is always the project-progress/ folder inside each project.
Path | What lives there |
| Canonical Markdown templates for new projects |
| Cross-tool skill instructions for maintaining progress |
| Reusable AGENTS.md, Claude Code, and hook guidance snippets |
| TypeScript progress validation and lifecycle hook checks (compiled to |
| PowerShell and POSIX wrappers that run the compiled hook on Node |
| TypeScript MCP server over project-local progress files |
| The Claude Code plugin |
| The Codex plugin (same skill and |
License
MIT — see the license field in package.json.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections. Dates show when Glama detected each change.
No tool schema history has been recorded yet.
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Connectors
Nifty's MCP server — exposes tasks, projects, messages, and files as tools for AI agents.
MCP Server for an Agent Task Marketplace
MCP server for generating rough-draft project plans from natural-language prompts.
Related MCP Servers
- FlicenseNot gradedqualityDmaintenanceA lightweight MCP server that enhances AI agents with tools for codebase analysis, task delegation to sub-agents, multi-agent coordination through chatrooms, and project todo management.-
- AlicenseBqualityDmaintenanceExperimental MCP server for coordinating CLI agents across projects using file-based task boards and assignment tracking.51MIT
- AlicenseNot gradedqualityDmaintenanceMCP server that allows AI agents to read and update project progress across different AI coding tools and sessions, maintaining a single source of truth in a PROGRESS.md file.17MIT
- AlicenseNot gradedqualityDmaintenanceEnables AI agents to manage hierarchical tasks, track progress, handle dependencies, and coordinate work through an MCP server.3615GPL 3.0
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/AndriiLavrekha/awesome-progress-tracker'
If you have feedback or need assistance with the MCP directory API, please join our Discord server