Eclusia
Freemium
Legal professionals handling confidential cases
An AI agent approves a pull request, then quietly swaps the payload before the connector fires — and you never know until prod breaks. Axtary exists to close that gap: it binds every approval to the exact payload hash reviewed, so a changed diff is a rejected diff.
Axtary sits beside the agent and evaluates each tool call against deterministic policy before the underlying provider is called. Routine actions — ones that clear the policy rules — pass automatically. Higher-risk actions pause for a human to review the normalized payload, and the approval is cryptographically tied to that specific hash. If anything changes after sign-off, the adapter catches the mismatch and blocks execution. Every attempt, pass, and rejection lands in a verifiable ledger. The self-hosted path keeps provider credentials local, which matters for teams that cannot route credentials through a third-party service.
Bottom line: Pick Axtary for a pilot where you need auditable, payload-level approval gates on agent-driven GitHub or Slack actions — but expect to build your own integration scaffolding if your stack falls outside the documented connector set.
Run the published CLI and packages beside an agent in your own environment.
Add shared review and evidence for one agreed non-production workflow after a 30-day pilot.
For requirements that need a written deployment, identity, retention, or support scope.
View full pricing on axtary.com →
Pricing may have changed since last verified. Check the official site for current plans.
Find your fit
More options in Guardrails & Safety. Compare pricing, platforms, and features side by side.
Freemium
Legal professionals handling confidential cases
Free · Open source
AI security researchers
Most authorization layers tell an agent what tools it can use. They say nothing about what it is allowed to do with those tools on a specific payload at a specific moment. Axtary intercepts the tool call, normalizes the action record — actor, task, resource, constraints, and payload hash — evaluates it against policy rules, and either passes it, blocks it, or routes it to a human for review of that exact payload. The signed ActionPass artifact expires and is bound to the reviewed hash; the connector adapter verifies the hash again before calling the provider. If the payload has been altered between approval and execution, the call is denied and both hashes are recorded in the ledger.
The payload-binding mechanism is the differentiating feature. Approval is not a broad permission to invoke a tool — it is authorization for a specific action with specific content. The vendor’s demo illustrates this with a tampered Slack message that injected a credential token after human sign-off; the adapter caught the hash mismatch and blocked the post before the Slack API was called. This closes the class of attack where an agent or compromised intermediary re-writes a payload after a human has already reviewed a summary rather than the literal content.
Axtary fits teams running pilots of agent workflows against security-sensitive targets: production repositories, infrastructure paths, internal Slack channels, cloud storage with sensitive prefixes. The ledger provides evidence for post-incident review and compliance documentation in non-production contexts. Where it does not fit: teams that need the tool to plan or orchestrate work — Axtary enforces policy on actions agents propose, it does not generate or sequence those actions. Teams whose connectors fall outside the documented set (GitHub, Slack, Linear, AWS/GCP reads, MCP tool calls, document search) will need to build adapter-side verification themselves before the hash-binding guarantee applies.
The self-hosted deployment runs policy evaluation locally so provider credentials never leave the host environment. The vendor describes a CLI-first workflow — `axtary run workflow` — used for reproducing and testing policy scenarios. ActionPass artifacts are designed to be consumed across SDKs, proxies, and MCP wrappers, so teams already using MCP-compatible tooling can slot verification in without replacing the agent runtime.
Sign in to rate this tool or suggest an edit.