MCP servers, AI backends,
and deep integrations.
MCP servers, skill snippets, and AI backends. Plug in the tools your team already uses with one-click installs.
Built for Claude Code.
Ironspire is optimised for Claude Code with full CLI and API backend support, per-agent model selection, and session resume. It also ships with native support for 11 additional providers, so every agent on your team can run on the model that fits best.
CLI backend with session resume, tool approval gating, and plan mode. API backend for direct access with streaming and exact cost tracking. Assign Opus, Sonnet, or Haiku per agent.
OpenAI, Gemini, Bedrock, Mistral, DeepSeek, Grok, Groq, Together, Fireworks, and Ollama. Assign any provider and model to any agent. Switch in seconds, conversations preserved.
Connect any OpenAI-compatible API: LM Studio, vLLM, LiteLLM, or your own inference server. Name it, point it at a URL, and it appears in the model picker alongside first-party providers.
The tools your team already uses.
One-click install from GitHub, Slack, Linear, Sentry, PostgreSQL, and dozens more. Category filtering, instant search, and per-preset recommendations.
- One-click install from the official MCP registry
- Per-preset recommendations surface the right servers
- Category filtering and instant search across 10,000+ servers
- Each server exposes typed tools agents can call directly
Fully customisable, every layer.
Presets, project rules, hooks, skills, cost gates, and approval flows. Every aspect of agent behaviour is configurable to suit your team's workflow.
- Agent presets with curated tool sets, models, and system prompts
- Project-level CLAUDE.md rules injected into every session
- Lifecycle hooks for pre-commit checks, test runs, and deploys
- Custom slash commands, skills, cost alerts, and approval gates
- All fully customisable per agent, per project, or globally
Every agent, any backend.
Assign each agent its own provider and model. Claude Code CLI and API are first-class, with native support for OpenAI, Gemini, Bedrock, Mistral, DeepSeek, Grok, and more.
- Claude CLI backend with session resume, tool approval gating, and plan mode
- Direct API access for every supported provider with streaming and per-turn cost tracking
- Per-agent model assignment: mix Opus, GPT-4, Gemini, and local models on the same team
- Custom OpenAI-compatible endpoints for self-hosted or proxy servers
Import your existing setup in seconds.
Ironspire auto-detects MCP servers, system prompts, and configurations from your existing tools. One-click import with conflict resolution, secret filtering, and full undo support.
VS Code
Cursor
Claude Code
CopilotYour repository, fully mapped.
Ironspire scans your git history and GitHub activity to auto-populate the Blueprint with milestones, detect commit patterns, and surface PR and issue context for your agents.
- Semver tag detection maps releases to Blueprint milestones automatically
- Conventional commit analysis surfaces feature, fix, and refactor patterns
- GitHub PR and issue sync with incremental updates and ETag caching
- Branch classification: active, merged, and stale branches at a glance
Paste an image, get an answer.
Drop screenshots, diagrams, or mockups directly into the chat. Agents with vision-capable models analyse the image inline, no file upload workflow needed.
- Paste images from clipboard or drag files into the chat input
- Native pass-through to Claude, GPT-4o, and Gemini vision APIs
- Use for UI reviews, bug screenshots, architecture diagrams, and design feedback
flex-shrink-0 on the icon container. Here is the fix...”Every tool in one searchable catalogue.
The Ironspire Marketplace aggregates the Official MCP Registry, Smithery, npm, and community-published packages into a single browsable catalogue. Search across 10,000+ servers, filter by category, and install with one click. Every package carries a trust tier, install count, and publisher verification so you always know what you are adding to your stack.
Database access, queries, schema inspection
Error tracking and performance monitoring
Design tokens, components, and assets
Automated review with configurable rules
Secure credential and secret management
Database access, queries, schema inspection
Error tracking and performance monitoring
Design tokens, components, and assets
Automated review with configurable rules
Secure credential and secret management
One search bar queries every registry at once. Filter by category, trust tier, or popularity. Results surface the best match regardless of where the package is published.
Every package is assigned a trust tier: Official, Verified, or Unverified. Automated code scanning runs on submission and Sigstore signing provides keyless verification.
No config files to edit. Select a package, click install, and it is live in seconds. Dependencies resolve automatically and per-server cost tracking starts from the first call.