Subagent execution is capped at a depth of five.
Just landed nested subagent support in Claude Code
Starting to experiment more with agents kicking off agents as a way to better manage context. Capped at depth=5 to start, going out in today’s release.
Lmk what you think!
Many users praise Claude Code's nested subagent support for enabling agent delegation and smarter context management, while others call for fixes to core model usability and token waste instead.
paging @kieranklaassen
Just landed nested subagent support in Claude Code
Starting to experiment more with agents kicking off agents as a way to better manage context. Capped at depth=5 to start, going out in today’s release.
Lmk what you think!
One idea I’m experimenting with: add "fork:true" to a skill’s frontmatter to have it run in its own context window, then in the skill tell it to use agents to keep context isolated for each step also.
Adding this to the built in /code-review skill to improve performance even more
“agents kicking off agents as a way to better manage context”
“nested” language models
“depth=N”
neat :-)
Just landed nested subagent support in Claude Code
Starting to experiment more with agents kicking off agents as a way to better manage context. Capped at depth=5 to start, going out in today’s release.
Lmk what you think!
@bcherny Testing this out as soon as I have access to it. Subagents are really underrated especially for effective context management.
Just landed nested subagent support in Claude Code
Starting to experiment more with agents kicking off agents as a way to better manage context. Capped at depth=5 to start, going out in today’s release.
Lmk what you think!
@omerfrkdemiral Say “use a workflow”. Just “workflow” had too many false positives
@bcherny Previously, when I wrote workflows, they were colorful; could you bring that back?
@bcherny like the context management angle
more agents kicking off agents = more review surface area
@plannotator feels timely
@bcherny just call it RLM already
@bcherny Now tracking what agents did and improving will get terrible.
But Superdense can help https://github.com/Nimrobo/superdense
@bcherny Can you share an example of how you personally use it?
@bcherny Does this include forked?
@bcherny Can we watch what each subagent is doing? Monitoring subagents should be fun!
@SeekerClaw Yep!
@bcherny @omerfrkdemiral To trigger built-in Claude features, it should be more obvious somehow. Maybe prefix them? c-workflow, c-loop, etc., when I want to explicitly trigger one inside my prompt.
Also, /workflow feels pretty awkward for opening the workflow overview.
@bcherny Thank you, owner of my dreams.
@bcherny testing depth=5 as a hard limit or more of a vibe recommendation?
@bcherny are they able to use chrome etc?
@bcherny Can we PLEASE get a “command view” of multiple agents at a time/threads. There no good way to see if I’m doing projects in parallel what the agents are doing.
How about you fix compaction instead like OpenAI did with CODEX? CODEX can compact and run one session for a week and still retain quality. This way /goal in CODEX works exceptionally well. It does not need to rely on subagents to keep context clean.
I can't explain it and don't know whats behind OAIs tech but this is fact. Their compaction mechanism is on another level.
You can be sure that CODEX mostly stays on track after compact. Right now one of my sessions (a coordinator / planner) has been running for almost 2 weeks already because im lazy to transfer context lmao
@bcherny Let me check then! At the same time was thinking about that the future will be prompt Engineer < Loop Engineer < Agent Engineer.
So you will create the Agent who has Access to the Web, Login and co and create they own agents. What is your thought?
Subagent execution is capped at a depth of five.
Just landed nested subagent support in Claude Code
Starting to experiment more with agents kicking off agents as a way to better manage context. Capped at depth=5 to start, going out in today’s release.
Lmk what you think!
