Complete reference for all 21 CodeSpar agent commands, including syntax, risk levels, and usage examples.

Command Reference

CodeSpar agents respond to commands via @codespar <command> in any connected channel (WhatsApp, Slack, Discord, Telegram). Commands can also be sent through natural language — see Natural Language Understanding for details.

Command Syntax

All commands follow the pattern:

@codespar <command> [arguments]

The @codespar mention prefix is required in group chats. In direct messages, you can omit it.

Risk Levels

Every command has an assigned risk level that determines whether it requires approval at a given autonomy level:

Risk Level	Color	Meaning
low	🟢 Green	Read-only or informational. Never requires approval.
medium	🟡 Amber	Modifies state or triggers work. May require approval depending on autonomy level.
high	🟠 Orange	Deploys or changes infrastructure. Requires approval at L0–L2.
critical	🔴 Red	Destructive or irreversible. Always requires approval (except L5 for non-production).

All Commands

Command	Syntax	Risk	Description
`status`	`status [build\|agent\|all]`	low	Query current status of builds, agents, or both
`help`	`help`	low	Show available commands and usage
`logs`	`logs [n]`	low	Show recent activity log (default: 10 entries)
`instruct`	`instruct <task>`	medium	Execute a coding task via Task Agent
`fix`	`fix <issue>`	medium	Investigate an issue and propose a fix
`review`	`review PR #<n>`	low	Review a specific pull request
`prs`	`prs [open\|closed\|all]`	low	List pull requests for the linked repo
`deploy`	`deploy [staging\|production]`	high	Trigger a deployment pipeline
`rollback`	`rollback [staging\|production]`	critical	Rollback the last deployment
`approve`	`approve [token]`	medium	Approve a pending action
`autonomy`	`autonomy L<0-5>`	medium	Set the agent autonomy level
`link`	`link <repo-url>`	medium	Link a GitHub repository to this project
`unlink`	`unlink`	medium	Remove the current project link
`context` / `memory`	`context` or `memory`	low	Show agent memory and vector store stats
`whoami`	`whoami`	low	Show your identity and connected channels
`register`	`register <name>`	low	Register or update your display name
`merge`	`merge PR #<n> [squash\|rebase]`	medium	Merge a pull request
`plan`	`plan <feature>`	medium	Break down a feature into sequential sub-tasks via Planning Agent
`lens`	`lens <question>`	low	Query data and get insights
`demo`	`demo [mcp-generator\|query <question>]`	low	Interactive feature demos
`kill`	`kill`	critical	Emergency kill switch -- stops all agent activity

Command Details and Examples

status

Query the current status of the linked project, builds, and agents.

@codespar status

Example output:

📊 Project Status
─────────────────
Project: codespar/codespar
Branch: main
Last build: ✅ passed (3m ago)
Active agents: 1 (Project Agent)
Autonomy: L1 (Notify)

With arguments:

@codespar status build

🔨 Build Status
─────────────────
Last commit: abc1234 — "fix: resolve auth token refresh"
CI: GitHub Actions
Status: ✅ passed
Duration: 2m 34s
Triggered: 8 minutes ago

@codespar status agent

🤖 Agent Status
─────────────────
Project Agent: ACTIVE (uptime: 4h 22m)
Task Agents: 0 running
Review Agents: 0 running
Deploy Agents: 0 running
Incident Agents: 0 running
Coordinator: IDLE

help

Display all available commands with brief descriptions.

@codespar help

Example output:

📖 CodeSpar Commands
─────────────────
status [build|agent|all] — Query current status
help — Show this message
logs [n] — Show recent activity
instruct <task> — Execute a coding task
fix <issue> — Investigate and fix an issue
review PR #<n> — Review a pull request
prs [open|closed|all] — List pull requests
deploy [staging|production] — Trigger deployment
rollback [staging|production] — Rollback last deploy
approve [token] — Approve pending action
autonomy L<0-5> — Set autonomy level
link <repo-url> — Link GitHub repo
unlink — Remove project link
context — Show agent memory stats
whoami — Show your identity
register <name> — Register display name
kill — Emergency stop

logs

Show recent activity entries from the audit log.

@codespar logs

Shows the last 10 entries by default. Pass a number to customize:

@codespar logs 25

Example output:

📋 Recent Activity (last 10)
─────────────────
[14:32] ✅ Build passed — main@abc1234
[14:30] 🔨 Build started — main@abc1234
[14:28] 📝 PR #42 reviewed — 2 suggestions
[14:15] 🚀 Deploy to staging — completed
[14:10] ✅ Approval received — deploy staging (by @john)
[14:05] ⏳ Approval requested — deploy staging
[13:50] 🔧 Task completed — "add rate limiting"
[13:45] 📋 Task started — "add rate limiting"
[13:30] 🔗 Repo linked — codespar/codespar
[13:28] 👤 User registered — John (slack:U1234)

instruct

Tell the Task Agent to execute a coding task. The agent clones the repo, makes changes in a Docker container, and opens a PR.

@codespar instruct add input validation to the /api/users endpoint

Example output:

🔧 Task Agent Started
─────────────────
Task: "add input validation to the /api/users endpoint"
Agent: task-abc123
Status: ACTIVE

I'll analyze the endpoint and implement validation.
Stand by for updates...

After completion:

✅ Task Completed
─────────────────
PR #47 opened: "feat: add input validation to /api/users"
Files changed: 3
Lines: +45 / -2
Branch: codespar/add-input-validation

Changes:
• Added zod schema for user input
• Added validation middleware
• Added tests for invalid payloads

fix

Investigate an issue and propose (or apply) a fix.

@codespar fix the auth token refresh is failing with 401 errors

Example output:

🔍 Investigating Issue
─────────────────
Issue: "auth token refresh failing with 401 errors"
Agent: task-fix-def456

Analyzing error patterns...
Found: Token expiry check uses UTC but server returns local time.

📝 Proposed Fix:
• File: src/auth/token.ts (line 42)
• Change: Use dayjs.utc() instead of dayjs()
• Risk: low

PR #48 opened: "fix: use UTC for token expiry comparison"

review

Trigger a code review of a specific pull request.

@codespar review PR #42

Example output:

📝 PR Review — #42
─────────────────
Title: "feat: add rate limiting middleware"
Author: @alice
Files: 4 changed (+120 / -3)

Review Summary:
✅ Code quality: Good
⚠️ 2 suggestions:
  1. [src/middleware/rate-limit.ts:15] Consider using sliding window
     instead of fixed window for more even distribution.
  2. [src/middleware/rate-limit.ts:28] Missing error message in
     RateLimitError — clients will see empty 429 response.

🟡 Verdict: Approve with suggestions

prs

List pull requests for the linked repository.

@codespar prs
@codespar prs open
@codespar prs closed
@codespar prs all

Example output:

📋 Open Pull Requests (3)
─────────────────
#47 feat: add input validation — @codespar (2h ago)
#45 fix: resolve memory leak in worker — @alice (1d ago)
#43 chore: upgrade dependencies — @bob (3d ago)

deploy

Trigger a deployment to staging or production. Production deploys always require approval.

@codespar deploy staging

Example output:

🚀 Deploy Requested
─────────────────
Target: staging
Branch: main
Commit: abc1234

⏳ Waiting for approval...
Approve with: @codespar approve dp-a1b2c3
Expires in: 30 minutes

After approval:

✅ Deploy Complete
─────────────────
Target: staging
Duration: 1m 45s
URL: https://staging.example.com
Health check: ✅ passing

rollback

Rollback the last deployment. This is a critical-risk command and always requires approval.

@codespar rollback staging

Example output:

⚠️ Rollback Requested
─────────────────
Target: staging
Rolling back to: previous deploy (commit def5678)

⏳ Requires approval.
Approve with: @codespar approve rb-x9y8z7

approve

Approve a pending action. You can approve by referencing the token shown in the approval request, or approve the most recent pending action.

@codespar approve dp-a1b2c3

Example output:

✅ Approved
─────────────────
Action: deploy staging
Approved by: John (slack:U1234)
Executing now...

Without a token (approves the most recent pending action):

@codespar approve

autonomy

Set the agent's autonomy level from L0 (passive) to L5 (full auto). See Graduated Autonomy for full details on each level.

@codespar autonomy L3

Example output:

🔧 Autonomy Updated
─────────────────
Previous: L1 (Notify)
Current: L3 (Auto-Low)

Auto-executing: status, help, logs, review
Requires approval: instruct, fix, deploy, rollback

link

Link a GitHub repository to the current project agent.

@codespar link codespar/codespar
@codespar link https://github.com/codespar/codespar

Example output:

🔗 Repository Linked
─────────────────
Repo: codespar/codespar
Branch: main (default)
Webhook: ✅ configured
Stars: 1,234
Last commit: abc1234 (2h ago)

See the Link a Repository guide for prerequisites and details.

unlink

Remove the current repository link from the project agent.

@codespar unlink

Example output:

🔓 Repository Unlinked
─────────────────
Previous: codespar/codespar
Webhook: removed
Agent state: reset to IDLE

context / memory

Show the agent's memory statistics, including vector store usage and conversation history.

@codespar context
@codespar memory

Both commands are aliases and produce the same output.

Example output:

🧠 Agent Memory
─────────────────
Vector store: 1,247 embeddings
Storage used: 12.3 MB
Conversations: 89 messages (last 30 days)
Project context: 45 files indexed
Last sync: 15 minutes ago

whoami

Show your resolved identity across all connected channels.

@codespar whoami

Example output:

👤 Identity
─────────────────
Name: John Silva
Role: maintainer
Channels:
  • Slack: @john (U1234ABCD)
  • WhatsApp: +55 11 99999-0000
  • Discord: john#1234
Registered: 2024-01-15

register

@codespar register John Silva

Example output:

👤 Registered
─────────────────
Name: John Silva
Channel: slack (U1234ABCD)
You can now be identified across all channels.

kill

Emergency kill switch. Immediately stops all running agents and pending actions. This is a critical-risk command and always requires approval (except from emergency_admin role).

@codespar kill

Example output:

🛑 EMERGENCY STOP
─────────────────
All agents: STOPPED
Pending actions: CANCELLED (3 cancelled)
Deploy in progress: ABORTED

To resume: @codespar status
Agent will restart in IDLE state.

Next Steps

Natural Language Understanding — Send commands in plain language
Graduated Autonomy — Control what agents do automatically
Link a Repository — Connect your GitHub repo

Command Reference

On this page