Build, share, and install custom tools from the Sprigr marketplace. Connect your agents to any API, database, or external system, with a built-in IDE, V8 sandbox security, and one-click deployment.
A built-in IDE for writing and testing tools. V8 sandbox execution. Immutable versions with instant rollback. Four sharing tiers from private to certified. Tools compose with each other. Your agents gain capabilities without your team writing integration plumbing.
You focus on the work. Sprigr runs the paperwork.
A complete platform for creating, testing, deploying, and sharing tools that extend what your AI agents can do.
A Monaco-based code editor with syntax highlighting, autocomplete, and error markers. Define input schemas with JSON Schema, test with sample data, and deploy, all without leaving the dashboard.
Every tool runs in an isolated V8 sandbox, the same technology behind Cloudflare Workers. No file system access, no cross-tool interference, and network requests restricted to declared domain allowlists.
Tools can call other tools using tools.call(), enabling you to compose complex capabilities from smaller building blocks. Chain a weather check into a scheduling decision into a notification, each piece reusable independently.
Find tools by industry (trades, HVAC, plumbing), integration (simPRO, Xero, Gmail), or purpose (scheduling, invoicing, compliance). Combine filters across all dimensions to find exactly what you need.
Every deployment creates an immutable version. Installers can pin to a specific version for stability, update to the latest when ready, or roll back instantly if something breaks. Zero-downtime deployments.
Each tool is limited to 100 calls per minute per company. If an external API goes down, circuit breakers automatically stop retrying until it recovers. Protects your systems and external services from overload.
Sharing tiers, rate limits, and isolation built in from day one.
Private (your company only), Shared (specific companies you invite), Listed (public marketplace catalog), and Certified (platform-approved with a badge). Change the tier at any time.
Every tool is rate-limited at 100 calls per minute per company. Circuit breakers automatically stop calling a tool when its external dependency is failing.
Same technology behind Cloudflare Workers. No file system access, no cross-tool state, no way for one tool to interfere with another or the platform.
Three steps from idea to production tool.
Open the built-in Tool IDE from your dashboard. Write a handler function using a Monaco editor with autocomplete, syntax highlighting, and a live console. Define your input schema and declare which external domains your tool needs to reach.
Run your tool in a sandboxed environment with sample inputs and test secrets. Check the output, iterate on the logic, and deploy with one click. Each deploy creates a new immutable version with automatic rollback support.
Once deployed, your agents can call the tool during conversations and workflows. The agent sees the tool name, description, and schema, then decides when to invoke it based on conversation context. No manual wiring needed.
Every marketplace tool runs in an isolated V8 sandbox. A buggy or malicious tool cannot affect other tools, your agents, or the platform.
Read the security deep-dive →Yes. Every tool runs in an isolated V8 sandbox with no file system access and no ability to interact with other tools or the platform outside of its declared APIs. Network requests are restricted to domains you explicitly allowlist. Secrets are encrypted at rest and injected at runtime, they never appear in code, logs, or conversations.
Yes. Tools support four trust tiers: Private (your company only), Shared (specific companies you invite), Listed (public marketplace catalog), and Certified (platform-approved with a badge). You can change the tier at any time as your needs evolve.
Your tools can make HTTP requests to any domain you declare in the network allowlist. This includes REST APIs, GraphQL endpoints, webhooks, and any other HTTP-accessible service. Requests to domains not on your list are blocked at runtime.
If a tool throws an error, the agent receives the error message and can respond gracefully in the conversation. If a tool fails repeatedly (for example, due to a downstream API outage), circuit breakers automatically stop calling it for a cooldown period. Rate limiting prevents runaway loops at 100 calls per minute per company.
Yes. Every deployment creates an immutable version. If a new version introduces problems, installers can roll back to any previous version instantly from the app settings. You can also pin installations to a specific version to prevent automatic updates.
Extend your AI agents with tools built for your industry and your systems.