Spec-Driven Integration for Agents

Turn existing APIs into agentic capabilities. Declarative & open source.
Test drive Ikanos
Apache 2.0 · Production-ready

From API to agent context in three steps

A complete, production-ready loop on two Apache 2.0 engines.
1
Import
ikanos import openapi petstore.yaml
Turn an API you already own into a skeleton capability. Aggregate several upstream calls into one domain-shaped capability — not a 200-tool endpoint dump. No blank YAML file.
2
Lint
polychro lint capability.yaml
Validate deterministically with Polychro, in-process, sub-100 ms — at the speed an agent generates it. Drop it in a PR with the "Fix It" GitHub Action.
3
Serve
ikanos serve capability.yaml
Serve it with Ikanos as MCP, Skill, and REST from the same spec (plus a Control port for ops). One Docker image, built-in auth, OTel traces + Prometheus metrics.

Build the MCP server your agent needs — not the one your vendor ships.

The unit of match is the agent's use case, not any one vendor's API. A doc-review agent needs three tools from three systems — public Notion, private Jira, a Slack webhook — fused into one curated surface. No vendor server can produce that. One Ikanos capability can.
The others
Naftiko
Write integration code (LangChain, LangChain4j)
Declare a capability in YAML
Hand-write an MCP server (FastMCP, Spring AI)
Emit MCP from one spec
Generate a 1:1 200-tool mirror (Stainless, APIMatic)
Shape one domain capability
Consume a vendor's generic MCP server
Fuse many APIs into the agent's surface
Keep your runtime. Replace the hand-written glue. When the MCP transport revises, it's an engine bump — your YAML is unchanged.
Ikanos 1.0 Beta 1

What's new in Ikanos 1.0 Beta 1

  • First release stable enough for real integrations — after four hardening alphas.
  • Binary Content end-to-end: declare, serve, and document binary outputs in pure YAML across REST and MCP.
  • Native CLI fully fixed — pre-built binaries for macOS, Linux, and Windows.
  • Spec-Driven Integration: one YAML spec served as MCP, Skill, REST, and Control ports.
  • 100% Apache 2.0 — own the spec, not someone else's generic server.
Read the blog post

Before Naftiko

Cost | Waste + duplication
Teams rebuild what already exists because reuse is invisible. Manual catalog registration goes stale; 3rd-party API usage is fragmented across teams. Reuse metrics are activity-based, not outcome-based.
Risk | Ungoverned exposure
Shadow gateways, shadow MCP, and unknown integrations expand the surface area. Identity propagation and authorization across agent chains is unclear. Compliance and audit trails are incomplete.
Velocity | Slow + brittle delivery
Developers spend time on archaeology -- docs, broken links, reverse engineering. Specs and reality drift; changes are discovered too late. Copilots and agents lack trusted context and degrade as tool lists grow.

After Naftiko

Cost | Measurable reuse
Reuse is discoverable and measurable across domains. Existing APIs and specs become capability inventory that teams can actually use. Duplication drops because what exists is easy to find and evaluate.
Risk | Governed AI integration
Policy-driven discovery and runtime controls keep agents inside trust boundaries. Standardized MCP servers and Agent Skills are reviewable and auditable. Identity and access patterns are explicit and enforceable.
Velocity | Delivery compounds
Context is right-sized into capabilities, not raw provider complexity. Faster path from prototype to production through consistent packaging and documentation. Teams ship with confidence because discovery, constraints, and telemetry are built-in.

Start free. Stay free if you self-host.

Community (free) ships the two OSS engines plus free Crafter — the whole author → lint → run inner loop, self-hosted forever. Developer, Team, and Enterprise add managed hosting, governance (Warden), and fleet-scale operations (Skipper) when you need them. The same capability YAML survives every upgrade unchanged.

Read your enterprise the way Signals reads the Fortune 1000

Naftiko Signals is the open intelligence platform that reads the public footprint of every enterprise — public job postings, press releases, newsroom content, open-source activity — into a structured signal corpus. Fortune 1000 coverage substantially complete. 44 signal categories scored per company. Three analysis layers on top: AI Waves, Roles, Impact.

A community, a resource, and a company you can build with

Resources
Blog, newsletter, whitepapers, glossary
Explore
Ecosystem
Open source, ecosystem, discussions
Dive in
Careers
Join the crew from the start
Come aboard