Once a team adopts AI coding agents, it quickly becomes hard to see who is using them, how much, and whether development is actually getting faster. Rudel is an open source analytics platform built to fill that gap. Install the CLI on each developer’s machine and Claude Code and Codex session records are collected automatically, letting you review token usage, session length, model breakdown, and activity patterns across the whole team in one dashboard. It is developed by RUDEL. in Germany, and the code is published under the MIT license. You can use the hosted version at app.rudel.ai, or run the whole stack — ClickHouse included — on your own servers.
Key Features
- Sessions are collected just by installing the CLI: Setup takes three steps — install the package with npm, log in, and enable it. From then on, session records are uploaded automatically through a Claude Code hook, so developers never have to record anything by hand
- Supports both Claude Code and Codex: It is not limited to a single agent; OpenAI Codex sessions are handled in the same dashboard. Teams that use several agents in parallel avoid having their measurements split apart
- Session-level analytics dashboard: Token usage, session duration, share of usage by model, and activity time bands can be viewed per developer and per repository. The result reads as a team-wide trend rather than an individual work log
- ClickHouse-based analytics backend: Collected sessions are stored in ClickHouse, a column-oriented database, so aggregation stays responsive even as sessions pile up
- Designed for self-hosting: The full stack can be brought up in your own environment with Docker Compose. Organizations that would rather not hand source code to an outside service can keep the data under their own control
- Secret filtering before upload: Session records can contain prompts, source code, and command output, so the CLI detects known secret patterns and strips them before transmission. Worth noting: the official documentation itself states this is best-effort and not a guarantee
Pricing
| Plan | Monthly price | Key points |
|---|---|---|
| Cloud | $0 | Use app.rudel.ai as-is. No paid tier is listed |
| Self-hosted | $0 (MIT license) | Build it in your own environment with Docker Compose. Infrastructure costs are on you |
Pricing is as of August 2026. Please check the official site for the latest pricing.
Pros & Cons
✅ Pros
- Free and open source, so there is almost no cost barrier to getting started
- Self-hosting means the source code contained in session records never has to leave your organization
- The only work required of developers is installing the CLI; nothing needs attention in day-to-day use
- Claude Code and Codex are measured together, so metrics are not fragmented for teams using both
- The ClickHouse foundation helps keep aggregation responsive as session volume grows
⚠️ Cons
- Self-hosting requires the ability to operate Docker and ClickHouse, which is a burden for teams without infrastructure staff
- Because the mechanism collects session records themselves, you need internal agreement in advance on how secrets and customer code are handled
- It is a relatively new project, and the brand name and distributed package name appear to be in transition. Check the current state of the official site and GitHub before adopting it
- Supported agents center on Claude Code and Codex, so coverage is narrow for teams whose main agent is something else
- Per-developer usage becomes visible, which can be received as surveillance if the operating policy is handled poorly
Comparison with Similar Services
| Criteria | Rudel | ccusage | Anthropic’s official analytics | In-house BI aggregation |
|---|---|---|---|---|
| Main use | Team-wide session analytics | Local token and cost totals | Organization-level usage review | Freely designed metrics |
| Audience | Teams and organizations | Individuals | Organizations (eligible plans) | Teams and organizations |
| Supported agents | Claude Code / Codex | Mainly Claude Code | Claude Code | Your choice |
| Self-hosting | Yes (Docker Compose) | Runs locally | No | Yes |
| Setup effort | CLI install + stack setup | Nearly zero | Admin settings only | High (requires building it) |
| Price | Free | Free | Included in the plan | Infrastructure and dev cost |
Who Is It For
- Engineering leads who rolled out Claude Code or Codex but cannot yet describe actual usage in numbers
- Technical managers who need to explain the cost-effectiveness of AI coding tools to management
- Organizations that cannot hand source code to an outside SaaS and want the analytics stack in-house too
- Teams that want to verify with data, not intuition, which models and usage patterns actually produce results
- Engineering organizations that prefer to run and extend an open source tool themselves
Summary
Rudel is one of the few open source options for making the actual usage of AI coding agents visible at the team level. Being free and self-hostable carries real weight for organizations that are cautious about sending code outside. At the same time, since the mechanism collects session records, decisions about secret handling and internal agreement should be settled before adoption. A realistic path is to try the hosted version with a small team first, then move to self-hosting once you judge that it fits your operations.