Claude Code & Codex

TokenJam ships first-class support for the two most-used coding agents: Claude Code and OpenAI Codex. Both work with zero code changes. tj listens on OTLP and the agent emits its own telemetry.

Claude Code

Monitor every Claude Code session (costs, tool calls, API requests, errors) with two commands:

pip install "tokenjam[mcp]"
tj onboard --claude-code
# Restart Claude Code, then:
tj status --agent claude-code-<project>

tj onboard --claude-code does everything in one shot:

• Creates a shared config at ~/.config/tj/config.toml (one config for all projects).
• Writes OTLP exporter vars to ~/.claude/settings.json.
• Tags this project by writing OTEL_RESOURCE_ATTRIBUTES to .claude/settings.json.
• Registers the MCP server globally (claude mcp add --scope user tj -- tj mcp).
• Installs a background daemon (launchd on macOS, systemd on Linux).
• Adds Docker harness-compatible OTLP env vars to ~/.zshrc.

Claude Code must be restarted after running tj onboard --claude-code.

Adding more projects

Run once per project directory:

cd /path/to/other-project
tj onboard --claude-code   # tags this project, no reinstall needed
# Restart Claude Code

Each project gets its own agent ID (claude-code-<repo-name>), all sharing one server and one ingest secret.

MCP server

The MCP server gives Claude Code direct access to your observability data inside the session. 13 tools available after restart:

Tool	What it does
get_status	Current agent state: tokens, cost, active alerts
get_budget_headroom	Budget limit vs spend
list_active_sessions	All running sessions across agents
list_agents	All known agents with lifetime cost
get_cost_summary	Cost breakdown by day / agent / model
list_alerts	Alert history with severity filtering
list_traces	Recent traces with cost and duration
get_trace	Full span waterfall for a trace
get_tool_stats	Tool call counts and average duration
get_drift_report	Drift baseline vs latest session
acknowledge_alert	Mark an alert as acknowledged
setup_project	Configure a project for TokenJam telemetry
open_dashboard	Open the web UI (starts tj serve if needed)

The MCP server opens DuckDB read-only, so there are no lock conflicts with tj serve.

Per-project tagging

After installing globally, ask Claude Code:

“Set up TokenJam for this project”

Claude calls setup_project, which writes .claude/settings.json with the right OTEL_RESOURCE_ATTRIBUTES for this project.

Codex

Monitor every Codex session. Run once, globally:

pip install "tokenjam[mcp]"
tj onboard --codex

tj onboard --codex is project-agnostic. It writes to ~/.codex/config.toml (Codex’s single global config), so you only run it once, not once per project. Codex hardcodes service.name="codex_exec" in its binary, so all sessions appear under the same agent ID regardless of which repo you’re working in.

tj onboard --codex:

• Writes an [otel] block and [mcp_servers.tj] to ~/.codex/config.toml.
• Registers the MCP server so Codex can call TokenJam tools directly.
• Installs the background daemon (launchd / systemd).

Codex must be restarted after running tj onboard --codex.

tj status --agent codex_exec   # check it's working

The same 13 MCP tools available to Claude Code are available to Codex after restart.

Uninstalling

# Remove all TokenJam data, config, daemon, MCP registration, and env vars:
tj uninstall --yes

# Then remove the package:
pip uninstall tokenjam -y

tj uninstall cleans up everything set by tj onboard --claude-code: daemon, MCP server, ~/.config/tj/, OTLP env vars in ~/.claude/settings.json, OTEL_RESOURCE_ATTRIBUTES in every onboarded project’s .claude/settings.json, and the harness env block in ~/.zshrc.