Skip to content

Frequently asked questions

What is Ironspire?

Ironspire is a desktop development environment with built-in AI agent management. It has a full IDE (Monaco editor, terminals, file tree), and you can deploy specialist agents that work alongside you or run autonomously. Each agent gets its own context, tools and permissions, so work stays clean and traceable.

Each agent runs in its own sandboxed environment with configurable tools, permissions and context. You can run multiple agents simultaneously, each working on different tasks.

Is Ironspire free?

Ironspire has a free Starter tier that includes up to 3 agents, all 9 presets, the MCP plugin browser and basic monitoring. No credit card required.

The Forge tier includes unlimited agents, cross-agent memory, project orchestration (Blueprint), advanced analytics, proactive check-ins and agent relay. Forge is available monthly, annually, or as a one-time lifetime purchase.

See the pricing page for full details.

What operating systems are supported?

Ironspire supports Windows (10/11, 64-bit), macOS (Intel and Apple Silicon) and Linux (64-bit, .deb and AppImage). See the installation guide for download options and system requirements.

What AI models does Ironspire support?

Ironspire supports models from multiple providers:

  • Claude (Anthropic): Opus 4.6, Opus 4.5, Sonnet 4.5, Sonnet 4, Haiku 4.5, Haiku 3.5
  • OpenAI: GPT-4o, GPT-4.1, GPT-4.1 Mini, GPT-4.1 Nano, o3, o3 Mini, o4 Mini
  • Google Gemini: Gemini 2.5 Pro, 2.5 Flash, 2.5 Flash Lite, 3.1 Pro, 3 Flash, 3.1 Flash Lite
  • Mistral: Mistral Large 3, Mistral Small 4, Codestral, Devstral 2, Ministral 8B
  • DeepSeek: DeepSeek V3, DeepSeek R1
  • xAI Grok: Grok 4, Grok 4.1 Fast
  • AWS Bedrock: Nova Micro/Lite/Pro/Premier, Llama 4 Scout/Maverick, Llama 3.x series, Mistral (Bedrock)
  • Ollama: any locally hosted model (discovered at runtime)
  • Custom endpoints: any OpenAI-compatible API (Groq, Together AI, Fireworks AI, Azure OpenAI, etc.)

You can set the provider and model globally in Settings or override them per agent. See the model registry for the full capabilities matrix.

Do I need my own API key?

Each provider requires its own credentials, configured in Settings > Providers:

  • Claude (SDK mode): if you have a Claude Max or Claude Pro subscription, the SDK uses your CLI authentication. No separate API key needed.
  • Claude (API mode): requires an Anthropic API key from console.anthropic.com. You pay per token.
  • OpenAI, Gemini, Mistral, DeepSeek, xAI, Bedrock: each needs its own API key from the respective provider.
  • Ollama: runs locally with no API key required.
  • Custom endpoints: configured with a base URL and optional API key.

See the providers overview for setup details.

Where is my data stored?

All data is stored locally on your machine. Conversations, agent configurations, memory and settings live in your user data directory. Nothing is sent to Ironspire servers.

Local SQLite databases store cost history (per-profile, WAL mode), chat message persistence and GitHub intelligence data. The only external communication is between your agents and the AI model providers, which is a direct connection from your machine.

If you have an Ironspire account, your subscription status is synced with our server, but your conversations and agent data remain local.

Is Ironspire open source?

Ironspire is not open source. However, it is built on open standards:

  • MCP (Model Context Protocol): an open protocol for AI tool integration
  • Multi-provider model support: works with Anthropic, OpenAI, Google, Mistral, DeepSeek, xAI, AWS Bedrock, Ollama and any OpenAI-compatible API
  • Electron: open-source desktop framework

The MCP plugin system is open, and anyone can create and share MCP tools that work with Ironspire.

How many agents can I run at once?

  • Starter tier: up to 3 agents
  • Forge tier: unlimited agents

There is no hard technical limit on concurrent agents. Performance depends on your machine's resources and the AI model's rate limits. You can track costs per agent in the Analytics panel.

What IDEs can I import settings from?

Ironspire can import system prompts, tool configurations and project rules from 10 development environments:

  • VS Code, Cursor, Windsurf, JetBrains IDEs
  • Claude Code, Cline, Aider, Continue.dev, GitHub Copilot
  • AGENTS.md (repository-level agent instructions)

See the IDE import guide for details on what gets imported from each source.

Does Ironspire support Docker sandboxing?

Yes. You can toggle Docker sandboxing per agent in the agent configuration modal. When enabled, the agent's file system access and shell commands run inside a Docker container rather than on your host. Docker Desktop must be installed and running.

See the Docker sandboxing guide for setup and configuration.

Can Ironspire analyse my Git history?

Yes. Git Intelligence scans your local repository for commit patterns, contributor activity and file change frequency. GitHub Sync extends this with remote data (pull requests, issues, CI status) when you connect a GitHub token.

See the Git Intelligence overview for details.

Do agents support image and vision input?

Yes. You can drag and drop images or use the file picker to attach them to any conversation. Vision input works across all providers and models that support it. The model picker shows a vision indicator for compatible models.

Are chat messages saved across restarts?

Yes. Ironspire persists chat messages in a local SQLite database, capped at 200 messages per agent. When you relaunch the app, conversations are restored with session dividers marking where each session began. Older messages beyond the cap are pruned automatically.

Can I use a beta update channel?

Yes. In Settings > About, toggle on the beta channel to receive pre-release builds. Beta updates follow the same auto-update flow (guarded, force, or deferred) but include features and fixes ahead of the stable release.

Can I publish my own marketplace content?

Yes, with the Forge tier. You can create and publish MCP plugins, agent presets and prompt templates in the .isp package format. Published content appears in the Ironspire Marketplace for other users to install.

See the publishing guide for the full workflow.

Can agents communicate with each other?

Yes, in multiple ways. With the Forge tier:

  • Corpus engine (cross-agent memory) lets agents share observations and reflections. Agents write findings to a shared knowledge store with SQLite + FTS5 search, and other agents receive relevant memories automatically.
  • Relay engine (agent communication) provides direct agent-to-agent messaging via MCP tools. Use @mention routing to send messages between agents with loop prevention and audit trails. Agent messages appear in the activity feed.
  • Foundry engine (Blueprint orchestration) lets you define projects as milestones, phases and tasks that are delegated across agents with validation gates. Lifecycle decisions route to your approval queue.

What happens when the context window fills up?

Ironspire includes automatic context compaction. When an agent's context reaches 80% of its window (configurable from 50-95%), older messages are summarised to free space. The last 4 user messages are always preserved.

You can also trigger compaction manually with the /compact slash command. After compaction, a summary shows how many tokens were freed and the cost of compression.

See the compaction guide for details.

How do I stop all agents at once?

Press Ctrl+Shift+E (Cmd+Shift+E on macOS) to trigger an Emergency Stop. This immediately halts all running agents with no confirmation dialog. Agent statuses update instantly in the UI, then terminals are killed and backends cleaned up.

You can also right-click the system tray icon and select "Emergency Stop".

See the E-STOP guide for the full shutdown sequence.

How do auto-updates work?

Ironspire checks for updates via GitHub Releases on startup (after 5 seconds) and then every 4 hours. When an update is available, you can choose how to install it:

  • Guarded: waits for all agents to finish before installing
  • Force: stops agents immediately and installs
  • Deferred: installs once agents finish naturally

After updating, a What's New modal shows the changelog for the new version.

What does the system tray icon show?

The tray icon tooltip reflects your fleet status: "Ironspire" when idle, "Ironspire: N agent(s) running" when active, or "Ironspire: N agent(s) crashed" when there are issues.

Right-click the icon for quick actions including Show, Start/Stop All, Mute Notifications, Emergency Stop and Quit. By default, closing the window minimises to the tray rather than quitting.

How do I report bugs?

You can report bugs through:

You can also visit the community page for all support channels and resources.

When reporting a bug, include your Ironspire version (found in Settings), your operating system and steps to reproduce the issue.

Does Ironspire collect telemetry?

Telemetry is opt-in only. If you choose to enable it, Ironspire sends anonymous usage events via PostHog using a random UUID (no personal information). A session heartbeat pings every 30 minutes. You can disable telemetry at any time in Settings.

PreviousBlueprint Overview