KOI-net Protocol
Federated Protocol
Community
Links
Details
License Open source — MIT License (Python reference implementation at github.com/BlockScience/koi-net)
Dev Status 🟢 Active
Dev Status Detail Beta / Active development (KOI-net beta release October 2025; multiple production deployments including BlockScience's internal KOI, Metagov's KOI Pond, and RMIT-Telescope; specification continues to evolve through working group review and inter-organizational testing)
Owner BlockScience (primary steward); developed in collaboration with Metagov (Metagovernance Project) and RMIT University via the Telescope project; reference implementation at github.com/BlockScience/koi-net; lead developer Luke Miller (BlockScience Research Engineer); Senior Research Engineer Orion Reed leads adjacent Project Interlay work
Country USA (BlockScience based in Albany, New York); collaborators in international Metagov network and Australia (RMIT)
Start Year 2025
Stack Python reference implementation (koi-net library); HTTP for inter-node communication; JSON for knowledge event payloads; RIDs (rid-lib) for object naming; specification language-agnostic and externally implementable
Funding BlockScience operational funding (private firm with internal R&D allocation); Metagov nonprofit funding for KOI Pond integration; Australian Research Council Future Fellowship for RMIT-side work; collaborative research without dedicated KOI-net-specific funding tranche
Last Investigated Jul 1, 2026
Federated Protocol Attributes
Origins Inter-organizational knowledge exchange (originated from BlockScience's recognition that organizational knowledge management systems should be able to federate across organizations without centralizing the knowledge or imposing a global ontology; emerged from collaborative research with Metagov on self-governance of knowledge infrastructures and from BlockScience's prior work on knowledge networks and protocols)
Database Per-node knowledge graphs (each KOI node maintains its own knowledge graph using whatever storage backend the node operator chooses — typically a graph database or hybrid graph-document store); KOI-net Protocol itself does not mandate any database technology
Query Language HTTP/REST endpoints defined by the protocol for inter-node knowledge exchange (event publication, subscription, retrieval, proxy operations); query semantics over the underlying knowledge graphs are node-implementation-specific
Data Formats Knowledge events as structured JSON payloads addressed via RIDs; HTTP request/response envelopes; specification at github.com/BlockScience/koi-net documents endpoint payload shapes
Collaborative Live Editing N/A — KOI-net handles asynchronous knowledge event exchange between nodes; collaborative editing of individual knowledge objects is application-layer concern within each node
Rich Text Editing N/A — federation protocol
Mobile Support Library-level (Python koi-net usable on any platform supporting Python; HTTP-based protocol implementable from any mobile platform); no dedicated mobile client
Web Support Yes — HTTP-based federation makes web integration straightforward; KOI nodes typically expose web interfaces (LLM-mediated chat being the most common pattern across BlockScience's KOI, Metagov's KOI Pond, and RMIT-Telescope deployments)
Native Apps No first-party native apps (KOI-net is a federation protocol consumed by KOI node software; downstream applications such as Project Interlay's adaptive interfaces and spatial canvas tooling integrate with KOI-net but are tracked separately)
Terms Open Source / Free — MIT License; nodes are self-hosted and self-governed
Funds Unknown — combined funding across BlockScience, Metagov, and RMIT operational budgets; specific KOI-net development allocation not separately published
Based On RIDs / Reference Identifiers (used as the addressing scheme for knowledge objects, users, and groups across federated nodes); HTTP (transport binding for inter-node communication); concepts from prior knowledge-network research at BlockScience; co-designed as part of the KOI suite
Content Addressing RID-based addressing of knowledge objects; references-vs-referents distinction enables multiple addresses for the same referent across nodes with different observations; not strictly content-addressing in the IPFS sense (RIDs are typed references rather than content hashes) but functionally similar for cross-node knowledge identification
E2EE Not currently a core protocol feature — knowledge exchange is between trusted federated nodes over HTTPS; organizational-level access control rather than end-to-end encryption between individual users is the architectural focus; future protocol versions may add E2EE patterns if use cases warrant
Signature TBD — specification in active development as of beta release; signature semantics for knowledge events expected to evolve through ongoing working group review; current implementation uses HTTPS transport-layer security as primary integrity mechanism
Permissions Per-node access control (each node defines its own permissions over its knowledge graph); inter-node permissions governed by the federation relationship between nodes (peer agreements rather than centralized policy); flexibility in community data curation enabled by RID-based dynamic organization
Semantic Web Compatibility Compatible at the data layer (RIDs embeddable as URIs, knowledge events expressible as RDF where useful) but KOI explicitly rejects the global-ontology move that traditional Semantic Web work assumes; KOI's architectural commitment is to organizations developing rich shared understanding without enforcing a universal world-model
Federation Model HTTP-based federation between independently-operated KOI nodes; each node governs its own knowledge graph, access policies, and federation choices; full nodes participate in the network as first-class peers, partial nodes participate with reduced capabilities for resource-constrained scenarios; proxy nodes mediate inter-organizational communication where direct peer connection is undesirable; explicitly distinct from ActivityPub-based federation patterns
Instance / Server Requirements Self-hosted KOI node software (Python implementation reference, but specification permits other implementations); each organization or community runs its own node with its own storage, governance rules, and access policies; HTTP-reachable endpoint for inter-node communication
Account Portability Knowledge object portability via RIDs (referents are referenced consistently across nodes through their RIDs); user identity and account portability across nodes not currently a primary protocol concern (KOI focuses on inter-organizational federation rather than user migration); Unknown for full account-portability semantics
Discovery / Relay Architecture Direct node-to-node federation with proxy nodes for indirect inter-organizational communication; no centralized relay or discovery service; node-level configuration determines federation peers and trust relationships; scalability discussed in February 2025 architecture overview
Server Authority Model Each node has full authority over its own knowledge graph; knowledge exchange across nodes is consensual and bidirectional; no central authority over the federation; "Beneficent Dictator" governance pattern for the KOI-net Protocol specification itself, with Luke Miller (BlockScience) as primary developer guiding contributions and working-group review
Protocol Maturity / Standardization Beta / Active research (beta release October 2025; v3 is current iteration with v1 and v2 as predecessors; specification evolving through BlockScience-led working group with Metagov and RMIT-Telescope collaboration; not yet submitted to formal standards bodies; governance follows "Beneficent Dictator" pattern; implementations currently concentrated among the three collaborating organizations)
Data Sovereignty High at organizational level (each node owns and governs its own knowledge graph; sharing decisions made per-relationship at the node level); user-level data sovereignty depends on the host node's policies; explicit architectural commitment to enabling "organizations to come to richer understandings of their common world without enforcing a universal world-model that all must share"