Repository navigation
feat(agents): c0mpute agent with sub-agents and /agents, Claude Code UX - #595
Merged
Merged
Conversation
moshcode agents c0mpute (and /agents c0mpute in the pit) opens a coding agent whose model runs on the c0mpute peer network. moshcode is the harness: text tool protocol (Infernet ignores tools), read/write/edit/bash/glob/grep/ ls/todo + task subagents in parallel, permission prompts with shift-tab modes, /agents list/new/show/edit/rm/run, @agent, Claude Code .claude/agents reused. Print mode with -p. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…, re-indented edits; ground the prompt in real file names Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Contributor
|
Review the following changes in direct dependencies. Learn more about Socket for GitHub.
|
ThreatCrush Security Scan8 finding(s) in the 10 file(s) this pull request changes. MEDIUM: 8
106 pre-existing finding(s) elsewhere in the repository — **HIGH/CRITICAL**: 8 | **MEDIUM**: 79 | **LOW**: 19Not introduced by this pull request. The full set is in the Security tab.
…and 86 more. Full results in the Security tab. Snippets are redacted; ThreatCrush never prints matched credential material. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
moshcode agents c0mpute(and/agents c0mputein the pit) opens a coding agent whose model runs on the c0mpute peer network, with the same UX as Claude Code, sub-agents included.c0mpute has no coding CLI, so moshcode is the harness: the loop runs locally, model calls go through
@profullstack/ai-providers'C0mputeProvider(Infernet job + SSE, house-model fallback,ik_live_billing).What you get
task(sub-agents). Infernet ignores the APItoolsfield, so calls are a text protocol (<tool name="read">{"path": "…"}</tool>) parsed leniently: attributes, bare strings, fenced JSON, a reply that is only the args object, unclosed blocks, stray}}.taskhands a self-contained job to an agent with a fresh conversation, its own prompt and tool list; severaltaskcalls in one reply run in parallel; sub-agents can't spawn sub-agents. Built-ins:general-purpose,explore,plan./agents: list (grouped project / user / Claude Code / built-in, shows overrides),new(drafted by c0mpute or by hand),show,edit($EDITOR; built-ins copy to the project first),rm,run.@agent <task>runs one directly..moshcode/agentsand~/.moshcode/agents; existing.claude/agentsload read-only with tool/model names mapped.y/always /n/ type an instruction). Shift-tab cycles ask → accept edits → plan.-ybypasses. Read-only shell (ls,cat,git status/diff/log,rg) never asks.● Read(…)/⎿lines, spinner, esc/^C interrupt,/model/models/todos/compact/clear/cost/key/cwd. Print mode with-p(or stdin).-C <dir>.Verified
test/c0agent.test.mjs) with a scripted provider: parser incl. the exact malformed output qwen3.5 9B produced live, stream filter, every tool, permission gating, parallel sub-agents (peak concurrency 2), definition priority/shadowing, config 0600./help,/agents,/agents show,/agents new --manualwizard → saved + listed, plan-mode badge.huihui_ai/qwen3.5-abliterated:9b, fleet key): "npm test fails. Find the bug, fix it, and run the tests" → read test.js + math.js, editeda - b→a + b, rannpm test(exit 0), reported. Two earlier live runs failed; their raw output drove the parser/edit fixes in the second commit.Privacy: prompts and tool output go to the peer node that takes each job; the banner and README say so.
🤖 Generated with Claude Code