Neura Market
Septim Agents Pack logo

Septim Agents Pack

Free

10 named Claude Code sub-agents (Atlas, Luca, Canon, Ember, Tally, Nova, Ward, Mira, Juno, Pip) covering planning, architecture, brand, marketing, finance, design, legal, customer, research, and coordination. Drop into `.claude/agents/`. *By [@septimlabs-code](https://github.com/septimlabs-code)*

FreeFree tier
Type
Open Source
Founded
2026
Company
Septim Labs
Visit Website GitHub (40 stars)

About Septim Agents Pack

Septim Agents Pack is a curated collection of 12 markdown-based sub-agents for Claude Code. Each agent is a single plain-text file containing a system prompt, a tool whitelist, and a cost limit. The agents handle common developer and startup tasks—triage GitHub issues, sweep for dead code, create bug-fix PRs, conduct brief research, search Obsidian notes, draft support replies, classify expenses, write newsletters, research prospects, generate standups, and plan feature sequences. Installation is a single unzip into ~/.claude/agents/; no login, no platform, no subscription. Call any agent by name with an @ mention, and Claude Code dispatches it. The entire roster is open-source and can be edited, version-controlled, and shared like any other file.

Key Features

Each agent is a single markdown file with system prompt, tool whitelist, and cost limit
12 agents: Triage, Sweep, Patch, Brief, Recall, Forge, Reply, Receipt, Cinder, Beacon, Standup, Bench
Zero-setup installation: unzip into ~/.claude/agents/
Call any agent by name using @mention syntax in your prompt
Agents run in parallel when tasks overlap, share context when needed
No login, no SSO, no environment file required
Plain-text files are editable, version-controllable, and inspectable

Pros & Cons

Pros
  • Free and open-source; no subscription or platform lock-in
  • Transparent markdown files that are easy to read and edit
  • Each agent has a clear tool whitelist and cost limit for safety
  • No installation beyond copying a folder; works immediately with Claude Code
  • Agents can be dispatched in parallel for complex multi-step jobs
  • Fully version-controlled like any other project file
Cons
  • Requires Claude Code (paid Anthropic subscription) to run
  • Only functional within the Claude Code environment
  • Agents are pre-configured; customization requires manually editing markdown files
  • No built-in scheduling or automation; must be called from a Claude Code session

Best For

Automated triage and labeling of new GitHub issuesCodebase cleanup: detection of dead code, unused exports, missing testsFixing bugs with branch creation, test writing, and PR submissionResearch summarization from specified documentation with citationsSearching and stitching answers from Obsidian vaultsSafe Postgres migration generation with dry-run and approval flowDrafting customer support replies in the user's voice (never auto-sends)Expense classification and duplicate charge detection from CSV files

FAQ

How do I install the Agents Pack?
Unzip the folder into ~/.claude/agents/. No login, no SSO, no environment file. Confirm with "/agents" from any Claude Code session.
How do I call an agent?
Mention the agent name in your prompt using the @ symbol, e.g., "@triage handle issue #404". Claude Code dispatches the right agent with its tools.
Can I edit an agent's prompt or tools?
Yes. Each agent is a plain markdown file containing its system prompt, tool whitelist, and cost limit. You can edit it in any text editor and version control it like any other file.
Is there a subscription or recurring fee?
No. The Agents Pack is free (open-source). There is no subscription, no recurring fee. The Claude Code Mastery Bundle (which includes this pack) is a one-time payment if you want additional tools, but the agents themselves are free.