Runtime support audit

Agent runtime audit: snapshot of 25 September 2026.

What 11 agent runtimes documented about their hooks on 25 September 2026, scored against the AgentHook v0.2 tiers. This is a dated research snapshot, not a live ranking: runtimes change their hooks often, and formal tiers will only come from the conformance test harness.

The snapshot.

Which AgentHook lifecycle events each runtime documented on 25 September 2026, scored against the conformance tiers. Any runtime can move tier by exposing the missing hooks; the AgentHook specification itself does not need to change for them to do so.

Audit method: direct inspection of each runtime's official hook, plugin or callback documentation and changelogs, and of the public Agentic Thinking publisher repositories, re-audited on 25 September 2026 against the AgentHook v0.2 draft specification (previous review 19 May 2026). Tiers shown are preliminary indicators of what each runtime's documented hook surface makes possible, not test results; formal Bronze, Silver, and Gold tiers will be awarded by the AgentHook conformance test harness once it ships alongside specification v1.0. Silver needs matched hooks before and after each model call, so runtimes without them are Bronze-capable whatever else they expose. No runtime has yet shown Gold end to end: Hermes captures reasoning only when a publisher enables it, and retry correlation is undocumented elsewhere. The OpenClaw row describes its embedded and CLI runtime; its docs say not every harness emits every hook. Our six publishers were last updated on 6 May 2026 and do not yet use hooks added since. Runtime providers may self-attest by Pull Request to the AgentHook repository. This audit is for technical interoperability comparison, not procurement guidance. Snapshot date: 25 September 2026. We do not maintain this as a live ranking; runtime providers can submit updates by pull request.

Runtime
UserPromptSubmit
Pre / Post LLMCall
Pre / Post ToolUse
Reasoning capture
Tier
Status
AgentHook Conformance Fixturereference fixture for the AgentHook standard
yes
yes / yes
yes / yes
yes
Exemplar (not rated)
Reference fixture only
GitHub
HermesNousResearch
yes
yes / yes
yes / yes
yes (opt-in)
Gold-capable
Publisher shipped; conformance pending
GitHub
OpenCodeopen source
yes
yes / yes
yes / yes
yes
Silver-capable
Publisher shipped; conformance pending
GitHub
OpenClawembedded and CLI runtime
yes
yes / yes
yes / yes
no
Silver-capable
Publisher shipped; publisher forwards only part of this surface
GitHub
Gemini CLIGoogle
yes
yes / yes
yes / yes
no
Silver-capable
No HookBus publisher yet
OpenAI Agents SDKOpenAI
n/a (prompt set by caller)
yes / yes
yes / yes
yes
Silver-capable
No HookBus publisher yet
Claude CodeAnthropic
yes
no / no
yes / yes
no
Bronze-capable
Publisher shipped; conformance pending
GitHub
Codex CLIOpenAI
yes
no / no
yes / yes
no
Bronze-capable
Publisher shipped; conformance pending
GitHub
AmpSourcegraph
yes
no / no
yes / yes
yes (at turn end)
Bronze-capable
Publisher shipped; conformance pending
GitHub
CursorAnysphere
yes
no / post-only
yes / yes
yes
Bronze-capable
No HookBus publisher yet
GitHub Copilot CLIGitHub
yes
no / no
yes / yes
no
Bronze-capable
No HookBus publisher yet
Devin CLICognition
yes
no / no
yes / yes
no
Bronze-capable
No HookBus publisher yet

Microsoft Agent Governance Toolkit is tracked as a control and subscriber layer, not as an agent runtime publisher. Sources for every row are listed below.

Where each row comes from.

Official documentation and changelogs, all checked on 25 September 2026. Where the documentation did not say, we did not mark a capability as present.