Anthropic prompt caching for tool schemas and sessions
NousResearch/hermes-agent · cluster 20880 · agent
Data as of
Nothing open on GitHub. Fix merged in #76032. Check it covers your case.
The short answer
- Known?
- Yes. 5 related items grouped here, 1 issue already closed.
- Fix event
- A fix was merged to main in PR #76032. Check it covers your case. No release data is tracked.
- Tracker
- The maintainers' tracker marks this
resolved, deciding PR #76032. That is separate from the fix event. - Still open
- Nothing open on GitHub.
These items address caching of tool schemas and system prompts on Anthropic, including cross-session prefix caching, cache_control budget enforcement, and reducing input-token overhead for tool-heavy agents.
Open issues0
Open PRs0
Fixes on record
-
#76032 perf(prompt-caching): cache tool schemas on native Anthropic without history loss
files:
agent/agent_runtime_helpers.py,agent/auxiliary_client.py,agent/chat_completion_helpers.py,agent/conversation_loop.py,agent/moa_loop.py,agent/prompt_caching.py,run_agent.py,tests/agent/test_auxiliary_client.py,tests/agent/test_failover_identity.py,tests/agent/test_moa_aggregator_cache_control.py,tests/agent/test_prompt_caching.py,tests/run_agent/test_anthropic_prompt_cache_policy.py,tests/run_agent/test_run_agent.py
3 PRs closed without merging
From the close comments: superseded by another PR 2 · merged via another change 1
-
#27170 fix: enforce Anthropic cache_control budget · superseded by another PR →
#76032,#37611,#20880 -
#37611 perf(prompt-caching): cache tool schemas without history loss · merged via another change →
9fc12bf,#76032,#20880 -
#69633 fix(agent): support stable cross-session prefix caching on Anthropic · superseded by another PR →
#69704
Closed issues · 1
- Closed#20880Tool-heavy agents pay ~70% input-token overhead — `system_and_3` caching skips tools schema
How this cluster was grouped
-
Link evidence (
structural links throughout): every member is connected by structural evidence: shared changed files, or a closing/superseded reference. -
Pairwise link density
0.40: the share of item pairs that carry a link, counting a structural link in full and a relevance-only link at 0.36.1.00means every pair is linked.
Pick up a Lab task
Check whether a fix covers the open issues here, with your own agent. Bring back evidence for maintainers.
Open the LabAt a glance
- Subsystem
- agent
- Merged PRs
- #76032
- Closed issues
- 1
- Attempts
- 3 PRs closed without merging
- Fix link checked
- 1 Aug 2026