Ironspire: The IDE for AI Agent Teams
The IDE for AI agent teams.
Code, run agents, and ship products. One desktop app, no terminal required.
Free with your own API key. Forge from $15/mo.
Review the project configuration and coordinate with the team on deliverables
I'll review the project setup and check in with the team. Let me gather a status report.
Yes, it actually looks like this. Editor, terminals, chat, and live monitoring in one window.
Powerful tools, painful setup.
You spend an hour configuring MCP servers, write a CLAUDE.md that gets half-ignored, and by turn 40 the agent has forgotten what it was building. The tools are incredible. Getting them to work consistently is the actual job.
The terminal barrier
Claude Code lives in the terminal. Most people who would benefit from it never open one.
The configuration tax
MCP servers, CLAUDE.md files, memory systems, permission configs. Hours of setup before a single line of real work.
The context problem
One agent juggling your whole codebase. By mid-session the context is full of unrelated work and output quality tanks.
The coordination gap
You can open five terminals. They still can not talk to each other, share what they learned, or follow a plan together.
The whole stack, one window.
Use it as an IDE. Use it to run agents. Or hand off entire projects to an autonomous team. It scales to how you work.
A full IDE
Monaco editor, integrated terminals, file tree. Write and ship code without ever opening another tool.
An agent platform
Deploy specialised AI agents with roles, tools, and context. Run one or run a dozen, each focused on its own job.
Structured by default
Projects built with AI fall apart without structure. Ironspire tracks milestones, plans, and tasks so nothing gets lost between sessions.
One agent isn't enough.
Ironspire runs separate agents with their own context, tools, and roles. A coordinator distributes tasks across them. You see the finished work.
Focused context
Your developer agent sees code. Your QA agent sees tests. Nothing bleeds between tasks. Context stays clean.
Parallel execution
A developer and QA agent run at the same time on different parts of the codebase. No waiting.
Coordinated output
A coordinator routes tasks and collects results. You check the finished work, not every intermediate step.
Meet your team.
Every agent comes with the right system prompts, MCP tools, and context for its role. Pick a preset and hit run.
Project Manager
CoordinationTranslates business goals into technical specs, manages the roadmap, and coordinates agents.
Pre-configured, not blank slates
System prompts, tools, and workflows tuned for each role. Your agents are productive from the start.
Fully customisable
Change the system prompt, swap tools, adjust autonomy. Presets are a starting point, not a cage.
Add your own
Create custom agents for any role. Define the prompt, pick the tools, set the rules.
Skip the roleplay.
Ironspire was purpose-built to let creative people build cool shit, manage their project, and ship something that doesn't fall apart at scale.
Autonomy levels
Plan mode, normal, full access. Set it per agent. A senior developer gets free rein; a new agent gets guardrails.
Live analytics
Token count, cost, and context usage per agent, updated live. You know exactly what each agent is costing you.
One-click plugins
GitHub, Slack, Sentry, PostgreSQL, 100+ MCP servers. Browse, click, installed. No JSON files, no restarts.
Structured development
Milestones, phases, plans, tasks. Work is documented and traceable. When you come back Monday, you know where you left off.
Shared memory
What one agent learns, every agent can access. Facts, patterns, and context persist across sessions and restarts.
Profiles and workspaces
Each project gets its own agents, tools, and settings. Switch in one click. Nothing leaks between workspaces.
Built different.
Memory, relay, and orchestration. The three systems that turn separate agents into a team that actually works together. Forge plan only.
Cross-Agent Memory
Every agent records what it learns. Facts, patterns, and events flow into a shared memory feed. Search across all agents at once. The system spots cross-team patterns and surfaces them for you.
Learn more about Corpus →REST API uses cursor pagination
Endpoints return next_cursor token. Offset pagination deprecated in v3.
Always paginate before transform
Transforming before full pagination leads to partial data in reports.
Rate limit hit at 100 req/s
Triggered 429 errors in load test. Recommend exponential backoff.
Batch endpoint supports 500 items max
Documented in API v3.2 changelog. Previous limit was 100.
Developer + PM both reference pagination. Synthesized into team convention.
Agent Relay
Agents delegate work to each other through @mention routing. The PM assigns tasks, the Developer builds, the QA validates. Entire delegation chains run without you in the middle.
Learn more about Relay →Project Orchestration
Break work into milestones, phases, and tasks. Agents execute through a pipeline: research, brief, plan, execute, validate. Built-in gates ensure nothing ships unchecked.
Learn more about Foundry →4/4 tasks verified. Phase 03 complete, moving to Phase 04.
Up and running in three minutes.
Download & install
30 seconds
Single .exe installer. No config, no PATH edits.
Paste your API key
10 seconds
Anthropic API key or Claude Max. Stored locally, never sent to us.
Add an agent and start
60 seconds
Pick a preset, hit run. Your first agent works in under a minute.
Start free. Upgrade when you need more.
Starter
3 agents
- Chat + terminal
- BYOK (CLI + API)
- MCP plugins
- File tree + editor
- Auto-updater
Forge
Unlimited agents
- Everything in Starter, plus:
- Cross-agent memory (Corpus)
- Blueprint orchestration (Foundry)
- Agent relay (Relay)
- Full analytics
- Proactive check-ins
Team
Unlimited agents
- Everything in Forge, plus:
- Cross-machine sync
- SSO/SAML
- Admin dashboard