AgentTrail
AgentTrail provides observability and guardrails for AI coding agents. It records agent sessions and costs, detects repeated failures, backtests policies, and blocks, warns on, or holds risky actions before they affect a developer's systems.
At a Glance
- Solo builders and side projects
- Software development teams using AI coding agents
- Scaling organizations needing oversight and reporting
- Organizations with compliance, security, and procurement requirements
AI Tools by AgentTrail
(1)
AgentTrail Guard
Guardrail CLI for AI Coding Agents
Discussions
No discussions yet
Be the first to start a discussion about AgentTrail
Latest News
CLAUDE.md vs hooks vs AgentTrail: how to stop Claude Code ignoring your rules
AgentTrail Guard v0.4.1 released
“It's only dev data”: the agent that reached for --force-reset
OWASP Top 10 for Agentic Applications: a field guide for coding-agent teams
Products & Services
Free, open-source Apache-2.0 CLI installed as @agenttrail/guard. It runs locally as hooks for Claude Code, Cursor, and Codex CLI, checks commands and file changes against 74 inspectable guardrails, and can block, ask for approval, or warn before an action runs.
Hosted observability and governance product for solo developers and teams. It traces sessions using OpenTelemetry, detects repeat failures, suggests policies, backtests them on history, enforces them across developers, and provides approvals and audit exports on higher tiers.
An open rule library used by Guard and OS, with 74 curated policies across 11 packs covering destructive actions, secrets, weakened tests, and other risky coding-agent behavior.
Market Position
AgentTrail positions itself as an independent governance layer rather than another coding-agent vendor. Guard differentiates from CLAUDE.md and other rules files, coarse permission rules, and hand-written hooks by enforcing deterministic, inspectable policies outside the model; OS adds hosted session observability, repeat-failure detection, backtesting, team synchronization, approvals, and audit evidence. The official comparison describes Guard as local, free, open source, and privacy-first, while OS is hosted and team-oriented.
Leadership
Founders
Samuel Corso
CEO of AgentTrail and a technology professional with more than 10 years in tech and a focus on AI and fintech. His public GitHub profile links him to AgentTrail and lists Dunellen, New Jersey; he also identifies as CEO of AgentTrail on X.
Executive Team
Samuel Corso
Chief Executive Officer
Technology professional with more than 10 years in tech and a stated focus on AI and fintech; public profiles link him to AgentTrail and show him maintaining its Guard repository.
Loveneesh Dhir
Founding GTM
Growth, partnerships, ecosystem, and developer-programs operator. His portfolio describes prior work with Union Labs, Shardeum, Hashed Emergent, OG Club, Commudle, and Scaler, including GTM and community-building roles.
Founding Story
AgentTrail says it was born inside a software studio that runs AI coding agents on real client work. The initial vision was to give engineering teams a record of what agents did and enforceable guardrails for what they do next, so recurring agent mistakes become prevention rather than repeated manual fixes.
Business Model
Revenue Model
AgentTrail OS uses seat-based subscriptions with included governed-action allowances and usage overage charges; Enterprise uses custom pricing. AgentTrail Guard is free and open source.
Pricing Tiers
1 seat and 1 agent platform; 2,000 governed actions per month; 7-day retention; 74 curated policies; read-only after the action cap.
3,000 governed actions per seat per month included; $1 per additional 100 governed actions; 30-day retention, synced team enforcement, backtesting, and email support.
5,000 governed actions per seat per month included; $1 per additional 100 governed actions; 90-day retention, approval SLA tracking, audit export, personal policies, and priority support.
Custom governed-action allowance, volume overage pricing, custom retention/compliance/security/deployment options, dedicated CSM, and SLA.
Target Markets
- Solo builders and side projects
- Software development teams using AI coding agents
- Scaling organizations needing oversight and reporting
- Organizations with compliance, security, and procurement requirements
- Prevent destructive database commands, force pushes, destructive SQL, unsafe file deletion, and secret exposure
- Govern AI coding agents before commands or file changes reach local development systems
- Observe and audit coding-agent sessions, token usage, and cost
- Turn recurring agent mistakes and post-mortem findings into tested, enforceable team policies
- Provide approval workflows and evidence for teams deploying coding agents