Workmate for Coding Agents
Turn an approved coding agent into a personal assistant across the work tools you already use.
- Source repo
- ZhixiangLuo/10xProductivity
- Stars
- ★ 487
- Last updated
- 4mo ago
- License
- MIT
- Primary language
- Python
- FA score
- 57/100 · Major gaps
At a glance
- How it runs
- Works with
- Portable with changesCodex · Claude Code (Partial support)
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- An engineer needs a source-linked search summary spanning Slack discussions, Jira tickets, and GitHub changes.
- Not a fit if
- Teams needing turnkey workflows that run without human supervision
- Users unable to configure a local coding agent and tool access
- Maintenance
- Commits in the last 6 monthslast commit 2026-07-13latest release v1.1.0 (2026-05-27)
- Source review
- 57/100 · Major gaps
What does this agent do, and when should you use it?
10xProductivity is a local-first stack for building a personal work assistant around your existing coding agent, browser sessions, desktop apps, and permissions. It includes recipes for more than 25 tool connections, enterprise search, and Cursor and Claude Code skills for setup, UI discovery, and workflow creation. You configure access to your tools, delegate tasks in natural language, and coach the agent in a laptop session as it searches, drafts, triages, or updates work. Early triggers and a local runtime handle polling, scheduling, state, reply logs, and coding-agent invocation; documented commands include 10x-host and 10x-standup-prep. Credentials, private recipes, and runtime state are kept outside the repository by default in ~/.10xProductivity/, making this a fit for people willing to configure and coach workflows rather than seeking a turnkey hosted service.
Users configure access to Slack, Jira, GitHub, Confluence, and other systems through setup.md and recipes in tool_connections/; a connection may use an API, CLI, browser surface, or local files. The enterprise-search workflow searches connected tools for a topic, synthesizes an answer, and links to source material. Users can also coach the agent through tasks such as Jira sprint triage, stand-up preparation, and PR-description drafting, then capture repeatable methods as skills or workflows. The early runtime can start triggers and workflows through 10x-host, handling polling, scheduling, deduplication, reply logging, and invocation of Cursor, Claude Code, or Codex. 10x-standup-prep generates stand-up preparation, with --post available for posting. The repository also documents an example of live web research through Google AI Mode.
- An engineer needs a source-linked search summary spanning Slack discussions, Jira tickets, and GitHub changes.
- An engineering manager wants to find stale Jira sprint tickets and have the agent draft follow-up comments for review.
- A developer responding to a PR review request needs a summary that combines code, ticket, and CI context.
- A team member wants to summarize changes across Slack, Jira, GitHub, and a calendar, then coach the process into a trusted morning brief.
- An employee using an internal portal or custom company tool wants to teach the agent to access it through an API, CLI, browser, or local files.
Key setup facts
Pulled from the install and usage notes below. The project README has the current values.
- Install / run
pip install -e ".[dev]"
- Environment variables
TENX_PRIVATE_DIR- Config files
TENX_PRIVATE_DIR/.env
How do you install or deploy this agent?
Install Cursor, Claude Code, Codex, or another coding agent you trust, then clone the repository. If needed, follow setup-python.md to prepare Python and Playwright; create a virtual environment and install the development dependencies. Ask your agent to read setup.md to configure personal tool connections. Configure credentials required by each connection guide; the Slack polling example uses SLACK_XOXC and SLACK_D_COOKIE in TENX_PRIVATE_DIR/.env.
git clone https://github.com/ZhixiangLuo/10xProductivity.git
cd 10xProductivity
python3 -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"How do you use this agent?
After configuring at least one tool connection, ask your coding agent to run enterprise search:
Read workflows/enterprise-search/enterprise-search.md and search across my connected tools for <topic>.Once Slack polling is configured, start the local runtime with:
10x-host --trigger slack-polling --workflow workflows/assistant/assistant.md --engine cursorThe stand-up preparation example runs in dry-run mode first. Review the output and set the destination channel before adding --post:
10x-standup-prep --meeting-context "Daily stand-up for <team/project>" --dry-runWhat are this agent's strengths and limitations?
- Works through an existing local coding agent and the user's permissions, without requiring a new company-wide automation platform or Slack app.
- Provides recipes for more than 25 tool connections and a playbook for internal or custom tools.
- Skills for enterprise search, tool setup, workflow creation, and UI discovery can turn repeated tasks into reusable processes.
- Credentials and runtime state live outside the repository by default.
- Users must configure tool access, credentials, and the local runtime; the quick start involves several setup steps.
- Triggers, runtime, and reusable workflows are marked early; learning and durable memory remain on the roadmap.
- Automation depends on a local coding-agent session and its available browser, app, API, CLI, or file access.
- The repository warns that automating some external platforms may violate their terms of service.
How does this agent compare with similar options?
The README contrasts this approach with centralized automation platforms that require IT rollout: it works through the coding agent and permissions a user already has. It adds tool connections, skills, and workflows to Cursor, Claude Code, Codex, or Copilot rather than replacing those agents.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Workmate for Coding Agents This agent | 57 · Major gaps | CLI | ★ 487 | 4mo ago | Python | — |
| PipesHub Enterprise AI Context Platform | 65 · Some gaps | Self-hosted serviceFree + model costs | ★ 4.1k | today | Python | — |
| OwnPilot Personal AI Assistant | 70 · Some gaps | Self-hosted serviceFree + model costs | ★ 430 | 2mo ago | TypeScript | OpenAI API · Claude API |
| Okou | 54 · Major gaps | Hosted serviceFreemium | ★ 1.2k | today | TypeScript | OpenAI API · Claude API |
How does FollowAgents rate this agent?
Why each dimension lost points
The documentation emphasizes using the local machine and existing permissions, recommends keeping tokens, cookies, and runtime state under TENX_PRIVATE_DIR, and shows stand-up preparation in dry-run mode before review and --post. It also describes human supervision. This supports limited permission and external-action controls. However, the tool surface is broad, and the colleague-distillation instructions can read workplace data about colleagues and write personal profiles. The packet does not show a general least-privilege mechanism, sensitive-data retention policy, or rollback plan. The contribution guide asks for tested snippets, but this material does not establish source attribution or maintainer identity.
The README distinguishes available, early, and planned capabilities, and the sample tests align with some runtime interfaces, supporting partial consistency. Dependencies have version floors or pins, but the evidence does not establish availability, compatibility coverage, or dependency risk handling. The sample tests show limited error handling and do not establish comprehensive user-facing failure messages.
The README names developers, managers, product, support, analyst, sales, and operations scenarios and gives examples for search, stand-up preparation, and collaboration tools. It says the runtime is early and learning and memory are planned, and explains the intended separation between triggers and connections. Detailed capability boundaries and trigger conditions remain incomplete. Setup mentions Python 3.11, a virtual environment, and Playwright, but does not establish full operating-system or coding-agent compatibility.
The repository map, skill entry points, workflows, connections, and private-state layout are clearly indexed. Quick-start steps, command examples, and a Legal Notice link provide practical guidance. The MIT license is explicit. Skill and script names appear reasonably consistent, but the packet has no changelog or compatibility history; it links to contribution guidance without identifying specific maintenance responsibility or update commitments.
Enterprise search, connections, and stand-up preparation illustrate reusable outputs and potential cross-tool value; the README describes a path from coaching to trusted automation. Output quality remains dependent on user configuration and coaching. Setup and upkeep across many connections and personal skills are not quantified, and triggers and runtime are still early, so cost-benefit evidence is limited.
Some README claims about layout, setup, and capabilities are traceable to the supplied pyproject, tests, and skill entry points, and the tests contain a few concrete behavioral assertions. Corroboration across sources is limited and the packet is only a sample. Claims such as tool count, enterprise suitability, and operational reliability are not comprehensively substantiated here. Labels such as available, early, and roadmap help distinguish status, but some benefits and capabilities remain product claims rather than clearly separated validation results.
- This is a partial static sample; it cannot establish the contents of unseen files, runtime behavior, or controls not shown.
- Connections may access enterprise systems and act as the user; review the selected connection, permissions, and workflow instructions before enabling triggers or posting actions.
- The colleague-distillation skill involves workplace communications and personal traits; confirm authority to process that data and limit collection and retention carefully.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
v1.1.0 v1.1.0 — Personal assistant agent direction
Release notes excerpt
Summary 10xProductivity is evolving from a tool-connection library into a human-AI interaction platform for building personal assistant agents on top of the coding agents and tools you already use. What's changed - Reframed the root README around personal assistant agents, human-AI interaction, coaching, reusable skills, trusted workflows, learning, memory, and evidence-based self-awareness. - Moved the original tool-connection philosophy into tool_connections/README.md so the connection layer remains documented as the foundation. …View on GitHub ↗ -
v1.0.0 v1.0.0: A framework to automate workflows with coding agents
Release notes excerpt
Coding agents are not just for coding—they can automate real work across your tools, data, and workflows. 10xProductivity helps executives, managers, and ICs build secure automation systems today. No need for IT approvals, no additional permissions needed. All you need is a coding agent. The Three Pillars The open-source framework is built on: * 🤖 Coding Agents: Provide reasoning and execution. * 🔗 Tool Connections: Use your own identity and existing permissions to access the tools you already use. …View on GitHub ↗
FAQ
Do I need a new Slack app or an administrator-led deployment?
Which parts are ready to use, and which are still early?
Where are credentials and runtime state stored?
~/.10xProductivity/; set TENX_PRIVATE_DIR to use a different private directory.Will workflows automatically publish content?
--dry-run. Review its output and set a destination channel before adding --post.Is Workmate for Coding Agents open source?
Is Workmate for Coding Agents safe to use?
Is Workmate for Coding Agents still maintained?
Who makes Workmate for Coding Agents?
Thanks — we use this to improve the page.