agentwach Features & Capabilities
agentwach is built for engineers who have to debug what an agent did at 2 a.m. last Tuesday. Each capability is labeled with the plan it unlocks on.
15 features competitors don't ship together.
Tracing tools show you what happened. agentwach also tracks Copilot credits, caps per-user spend, pauses runaway loops, forecasts budget breach, and gates expensive models behind approval.
GitHub Copilot AI credit tracker
Per-user, per-repo, per-model Copilot premium-request credit tracking in real time.
Per-user AI spend limits
Cap daily or monthly spend per user across every provider. Block, throttle, or warn.
Per-repo Copilot credit reports
See which repos burn the most credits. Cap a repo before it hits 100%.
Agent kill switch
One-click stop for any agent. Returns stop=true on the next ingest — clean exit.
Auto-pause runaway loops
Detect 5+ repeat tool calls in 30s and pause the agent before it drains the budget.
Slack / Discord / Email action buttons
Pause, hard-stop, or raise the cap straight from the alert — no dashboard needed.
Cost forecast
“You will hit budget in 3 hours” — projected breach time on a rolling 5-min window.
Approval gates for expensive models
Require human approval before a session uses GPT-4o, Opus, or any model you flag.
Model efficiency report
Cost-per-resolved-task by model. Catch the moment a cheaper model would do the job.
Prompt-injection risk scanner
20+ heuristic rules block jailbreaks and exfiltration. Every block logged to audit.
Agency client workspaces
Isolated workspace per client. Per-client caps, keys, and exports — zero data crossover.
White-label client reports
Branded PDF and CSV exports your agency hands to clients each month.
Weekly AI spend report email
Monday digest — top users, top agents, top models, projected month-end overage.
“Money saved” dashboard
Counts the spend you would have lost to loops, runaways, and Copilot spikes.
Browser extension
ChatGPT, Claude, Gemini, and Copilot Chat captured into the same dashboard.
All capabilities by plan
Live agent roster
Every agent in your swarm shows up the moment it emits its first event. Status, current task, last activity, and run count update in real time.
Inter-agent message trace
A continuous feed of who said what to whom. Replay the conversation that led up to a failure without touching a log file.
Per-agent timeline
Drill into any agent to see every tool call, LLM call, state update, and error — chronologically, with full payloads.
Drop-in LangChain SDK
Add one callback handler to your LangGraph or LangChain pipeline. No re-architecting, no proxy, no waiting.
Workspace API keys
Rotate keys, name them per environment, see when each was last used. SHA-256 hashed at rest, never displayed twice.
Error-first surface
Agents in an error state are flagged at the top of the dashboard with the failing tool call inline. Stop scrolling through logs.
Token & cost monitoring
Track token usage and spend per agent, per model, per run. Bundled into every paid plan — no add-on required.
Extended retention & REST API
90-day event retention plus full REST API for exports and custom dashboards. Built for teams running agents in production.
Multi-workspace & SSO
Up to 10 workspaces, 2M events/mo, 1-year retention, dedicated support. SAML SSO available on Enterprise.
Tier label shows the lowest plan that unlocks the feature. See pricing for limits and overage rates.