- Home
- Skills
- Agents & Orchestration
- Agent Session Observatory
Agent Session Observatory
See every local Claude Code and Codex session, what it is doing, and what is waiting on you, on Mac, Windows, and Linux, plus any other agent's logs through a generic adapter.
$25
Works with the AI Agents you already use
Agent Session Observatory
Example session with this skill installed
Run a session summary for the last 48 hours and show me what is waiting.
- Read your context and instructions
- Compiled the agent session observatory
Found 4 active sessions in the last 48 hours. One session is currently waiting for your input.
Agent Session Summary
| Status | Agent | Project | Last Active | Context | Notes |
|---|---|---|---|---|---|
| WAITING | Claude | web-portal | 12 mins ago | 64% | Asks: "Should I proceed with deleting the deprecated /v1/auth endpoint?" |
| WORKING | Codex | data-pipeline | 2 mins ago | 12% | Now: Running pytest tests/test_ingestion.py |
| IDLE | Claude | docs-site | 5 hours ago | 5% | Last: Finished updating README.md |
| WORKING | Cursor | infra-aws | 1 min ago | N/A | Last text: "Scanning terraform state for orphaned resources" |
Current Fleet Status
- Waiting: 1
- Working: 2
- Idle: 1
Next steps
- Address the deletion prompt in the
web-portalproject terminal. - Check the
data-pipelinetest results in approximately 5 minutes. - Use
python3 -I collect-session.py --json --out report.jsonto export this data for your dashboard.
Connects securely to your tools. The creator never sees your data.
About this skill
Running multiple coding agents in parallel makes it impossible to track which terminal is stuck. This skill scans your local logs to surface sessions waiting for human input, ensuring you never leave an agent idling for hours.
What it does
- Identifies blockers by scanning logs for questions, approval requests, or decisions that require human intervention.
- Summarizes active sessions across Claude Code and Codex, including project name, title, model, and last activity timestamp.
- Monitors progress by reporting what each working session is doing right now, including its current task, last steps, and plan.
- Tracks resource usage by extracting context window percentages and rate limit windows from local metadata.
- Supports external agents through a generic adapter that parses any
.jsonlsession log from tools like Cursor or Copilot.
How it works
- Run
python3 -I collect-session.pyto print a summary of all active, waiting, and idle sessions to your terminal. - Use the
--hoursflag to adjust the lookback window for finding older sessions. - Invoke
--jsonto output session data for custom dashboards or internal tooling. - Set environment variables like
AGENT_OS_CLAUDE_DIRto point the scanner at non-standard log locations.
Frameworks & tools
Built with Python 3.11+ using only the standard library. It interfaces with Claude Code and Codex log files on macOS, Windows, and Linux.
Why this beats prompting it yourself
An agent cannot see what other agents are doing in separate terminal instances or processes. This skill operates at the OS level, reading the physical log files that LLMs cannot access themselves to provide a unified view of your entire agent fleet.
Use cases
- Find which of five parallel Claude Code tasks is waiting for a test suite approval.
- Check the remaining rate limit quota for Codex sessions before starting a large refactor.
- Monitor long-running agent tasks without constantly switching terminal tabs.
Known limitations
Cannot track agents that do not write local .jsonl logs. Requires Python 3.11 or newer. Context and rate limit data depend entirely on what the specific agent writes to its local metadata files.
Requires Python 3.11+. Only works with agents that write local .jsonl logs. Mac/Windows testing is pending; currently verified on Linux.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 8 days ago
Requires