Skip to content

Latest commit

 

History

History
198 lines (155 loc) · 8.95 KB

File metadata and controls

198 lines (155 loc) · 8.95 KB
systemprompt.io

Production infrastructure for AI agents

systemprompt.ioDocumentationCoreTemplate


Demo Suite

42 runnable demo scripts organized into 10 categories. Each demonstrates a different aspect of the platform via CLI commands.

Quick Start

# 1. Build and start services
just build && just start

# 2. Run preflight (acquires token, checks services, creates admin user if needed)
./demo/00-preflight.sh

# 3. Seed data for analytics and trace demos
./demo/01-seed-data.sh

# 4. Pick any category
./demo/governance/01-happy-path.sh
./demo/infrastructure/02-database.sh
./demo/analytics/01-overview.sh

Categories

Organized by the three pillars of systemprompt.io: Infrastructure, Capabilities, and Integrations.

Pillar Category Scripts What it covers Cost Recording
Infrastructure infrastructure/ 5 Services, database, jobs, logs, configuration Free infra-self-hosted, infra-control-plane
Infrastructure cloud/ 1 Auth status, profiles, deployment info Free infra-deploy-anywhere
Capabilities governance/ 8 Scope enforcement, secret detection, audit trails, hooks Free cap-governance, cap-secrets
Capabilities mcp/ 3 MCP server management, access tracking, tool execution Free cap-mcp
Capabilities analytics/ 8 Overview, agents, costs, requests, sessions, content/traffic Free cap-analytics
Capabilities agents/ 5 Agent discovery, messaging, tracing, A2A registry 2 @ ~$0.01 cap-agents
Capabilities users/ 4 User CRUD, roles, sessions, IP bans Free cap-compliance
Integrations skills/ 5 Skills, content, files, plugins, contexts Free int-cowork, int-extensions
Integrations web/ 2 Content types, templates, sitemaps, validation Free int-web-publisher
Integrations performance/ 2 Request tracing, benchmarks, load testing Free int-benchmark

Total: 43 category scripts + preflight + seed = 45 scripts. 41 free, 2 cost ~$0.01 each.

Prerequisites

  • Built binary: just build
  • Running services: just start
  • Preflight completed: ./demo/00-preflight.sh
  • Seed data (recommended): ./demo/01-seed-data.sh

Script Index

Setup

Script Description
00-preflight.sh Service health check, admin user creation, token acquisition
01-seed-data.sh Populate governance decisions, events, skills, and content

Infrastructure

Script Description
infrastructure/01-services.sh Service status and health checks
infrastructure/02-database.sh Tables, schema, queries, migrations
infrastructure/03-jobs.sh Job scheduling, execution history
infrastructure/04-logs.sh Log viewing, search, traces, requests
infrastructure/05-config.sh Full platform configuration overview
cloud/01-cloud-overview.sh Auth status, profiles, deployment info

Capabilities

Script Description
governance/01-happy-path.sh Governance ALLOWS admin-scope tool call
governance/02-refused-path.sh Governance DENIES user-scope agent
governance/03-audit-trail.sh Query governance decisions from database
governance/04-governance-happy.sh All 3 rules pass for admin agent
governance/05-governance-denied.sh Scope + blocklist deny for user agent
governance/06-secret-breach.sh Secret detection blocks credentials
governance/07-rate-limiting.sh Rate limit, security, and server configuration
governance/08-hooks.sh Hook listing and validation
mcp/01-mcp-servers.sh Server status, tools by server
mcp/02-mcp-access-tracking.sh OAuth + MCP tool call tracking
mcp/03-mcp-tool-execution.sh Tool listings, execution logs
analytics/01-overview.sh Dashboard overview
analytics/02-agent-analytics.sh Agent stats, trends, deep-dives
analytics/03-cost-analytics.sh Cost summary, breakdown, trends
analytics/04-request-analytics.sh AI request volume, latency, models
analytics/05-session-analytics.sh Session stats, trends, real-time
analytics/06-content-traffic.sh Content engagement, traffic, geo
analytics/07-conversations.sh Conversation stats, trends, listing
analytics/08-tool-analytics.sh Tool usage stats, trends
agents/01-list-agents.sh Agent discovery — admin and core views
agents/02-agent-config.sh Validation, MCP tool access, status
agents/03-agent-messaging.sh Full agent pipeline with AI (~$0.01)
agents/04-agent-tracing.sh Traces, artifacts, cost attribution
agents/05-agent-registry.sh A2A gateway, agent logs
users/01-user-crud.sh User listing, counts, stats, search
users/02-role-management.sh User details and role inspection
users/03-session-management.sh Current session, available profiles
users/04-ip-ban.sh Add/remove IP bans with verification

Integrations

Script Description
skills/01-skill-lifecycle.sh Skill listing, details, sync status
skills/02-content-management.sh Content listing, search, popularity
skills/03-file-management.sh File listing, config, storage stats
skills/04-plugin-management.sh Plugins, hooks, extensions, capabilities
skills/05-contexts.sh Context CRUD — create, show, edit, delete
performance/01-request-tracing.sh Typed data, flow maps, benchmarks
performance/02-load-test.sh 2000-request load test
web/01-web-config.sh Content types, templates, assets
web/02-sitemap-validate.sh Sitemap config, web validation

Connecting the Cowork desktop app

The enterprise-demo plugin, its skills, agents, and MCP servers ship with the template, but plugin assignment is per-usersetup-local populates the global registry and 01-seed-data.sh forks enterprise-demo into the active session's user. Whoever is authenticated when you run the seed gets the plugin.

Full flow to route Cowork through a locally-running template:

# 1. Bring the template up (from systemprompt-template)
just setup-local <anthropic_key> [openai_key] [gemini_key]
just start

# 2. Seed — this ALSO forks enterprise-demo for the current admin session.
./demo/00-preflight.sh        # admin session + /demo/.token
./demo/01-seed-data.sh        # + fork enterprise-demo for that user

# 3. Issue a PAT for the user you want Cowork to connect as.
#    Open http://localhost:8080/admin/devices → sign in → create device →
#    copy the sp-live-... PAT. (CLI PAT issuance isn't wired yet.)
#    If this is a different user than the seed target, fork for them too:
#      curl -X POST http://localhost:8080/api/public/admin/user/fork/plugin \
#        -H "Authorization: Bearer <their-JWT>" \
#        -H 'Content-Type: application/json' \
#        -d '{"org_plugin_id":"enterprise-demo"}'

# 4. Wire up the cowork helper (from systemprompt-core)
cargo build --release --manifest-path bin/cowork/Cargo.toml
BIN=bin/cowork/target/release/systemprompt-cowork
"$BIN" login sp-live-... --gateway http://localhost:8080
"$BIN" install --apply    # writes MDM keys for com.anthropic.claudefordesktop
"$BIN" sync               # pulls manifest → ~/Library/Application Support/Claude/org-plugins/

# 5. Cmd+Q Cowork and relaunch. It now reads inferenceProvider=gateway from
#    the MDM keys and routes every request through http://localhost:8080 with
#    a JWT minted by the helper.

Verify the gateway is wired:

"$BIN" status     # paths + gateway + last-sync timestamp
"$BIN" whoami     # identity JSON from the gateway

If sync reports 0 plugins / 0 skills / 0 agents / 0 MCP, the authenticated user has no forked plugin. Re-run the fork POST in step 3 with the correct user's token.

To revert Cowork to direct Anthropic: systemprompt-cowork uninstall then Cmd+Q and relaunch.

Troubleshooting

# If services are down
systemprompt infra services cleanup --yes
systemprompt infra services start --kill-port-process

# Wait for "All services started successfully" then retry

Recording

SVG terminal recordings and video recording infrastructure are in recording/. See recording/RECORDING-GUIDE.md for video production workflow.


License

MIT - See LICENSE for details.