Provides tools for analyzing OpenAI Codex task trajectories from local logs, including event timelines, turn/step summaries, token usage and cost estimates, and optional live stop controls.
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
English · 简体中文
See what a Codex task did, where time went, and how many tokens it used. Codex Trajectory turns local task logs into a searchable timeline and event ledger.
Replay tasks: inspect messages, saved reasoning summaries, tool calls, failures, subagent activity, and context compaction.
Find slow work: see durations and percentages for tools, estimated model time, and unattributed time in the selected task.
Track usage: view tokens, cache usage, and estimated costs, including all descendant subagents in task totals. Turn details show the selected agent only.
Follow live activity: watch new events, token totals, and the latest local quota sample.
Current viewer, captured October 22, 2026 with synthetic demo data.
Requires Codex desktop and uv. Supports macOS, Linux, and Windows.
codex plugin marketplace add icesixgod/codex-trajectory
codex plugin add codex-trajectory@icesixgodAfter installing, open a new Codex task. After updating, fully quit and reopen Codex.
To upgrade an existing Git marketplace installation, refresh the marketplace and install the current package:
codex plugin marketplace upgrade icesixgod
codex plugin add codex-trajectory@icesixgodThen fully quit Codex with Cmd+Q on macOS, or quit the app completely on Windows/Linux, and reopen the task's panel. Reopening only the panel can retain an older UI resource.
Related MCP server: agent-activity
In the task's right panel, select + → More tools… → Plugins and MCPs → Codex Trajectory. This opens the current task without a model turn; if task context is unavailable, choose a local task.
Search or filter the event ledger, then click a record to inspect it. Use Load earlier records for more history.
Click Calculate task timing for the timing breakdown, or Live window to follow new activity.
You can also open the viewer through show_codex_trajectory by asking Codex:
Show the safe trajectory summary for this Codex task.
For MCP calls, pass an explicit sessionId to select a task. Without it, read/show tools use the most recently modified active task.
Timing excludes gaps between turns and counts overlapping tools once. Model time includes network and queueing waits. Costs use bundled API prices and are estimates, not subscription charges; local quota samples may lag.
If a log exceeds 512 MiB, raise Read limit (GB) in the error panel and click Enable and retry. The limit applies only to that task in the current panel.
Logs are read locally and remain unchanged. Safe summary hides tool inputs, outputs, and raw metadata; Load full details requires confirmation and can remember your choice. Live views always use safe summaries.
Data returned through Codex may enter the active conversation. See the privacy policy for details.
MCP tools and API · Troubleshooting · Development · Changelog
MIT. Parts of the UI are adapted from DeepSeek Harness, © 2026 DeepSeek. See NOTICE and the upstream license. Independently maintained; not affiliated with DeepSeek.
Friendly link: Linux.do
This server cannot be deployed
An MCP server that provides read access to your cloud storage providers, bank accounts and more.
OAuth-protected, read-only-by-default MCP server for provenance-labeled QuillCaddie project memory.
Read-only MCP server for The Quiet Protocol's engines, benchmarks, proof, and business data.
The Cortex MCP server provides read-only access to real-time engineering context from the Cortex developer portal, allowing AI coding assistants to answer natural language questions about your organization's catalog (microservices, libraries, domains, teams, infrastructure), scorecards (engineering standards and best practices), initiatives (goals and deadlines), and Engineering Intelligence metrics. It includes tools for querying documentation, tracking personal entities, and accessing AI-assisted insights across the entire Cortex ecosystem.
