1
Repeated context is expensive
Large prompts, full repositories, tool schemas, and growing chat history consume the context window before the agent can make real progress.
Experimental · vendor-neutral · self-hosted direction
Context infrastructure for AI agents.
OCP AI is an experimental open context layer for AI agents. It helps agents retrieve only the context they need instead of repeatedly loading entire codebases into prompts.
95%estimated fewer repeated input tokens per agent step
27×more room for reasoning before context fills
Anymodel, IDE, or agent runtime direction
0vendor lock-in by design
01 · Why
AI agents are powerful, but context is still fragile.
Most agent workflows repeatedly rediscover the same project knowledge. They scan files again, ask for rules again, and lose useful state when the session ends.
2
Agents need memory boundaries
Context should be scoped by project, user, task, and team. That makes retrieval safer, easier to audit, and easier to reuse.
3
Knowledge should be portable
Teams should not have to recreate the same memory layer separately for every IDE, model, agent, or workflow tool.
02 · Solution
A portable context layer between agents and project knowledge.
Instead of sending full repositories into every prompt, OCP lets agents retrieve only the relevant context for each task.
Without OCP
Load repo→Resend history→Hit context limit→Repeat
With OCP
Index once→Retrieve chunks→Persist state→Continue
03 · How
The pipeline that makes code searchable by intent.
The first release direction is simple: index project knowledge, chunk it by meaning, retrieve what matters, and keep useful state available for future agent steps.
01 · Watch
Watch
Detect file and document changes without repeatedly reprocessing everything.
02 · Parse
Parse
Understand code structure at the level of symbols, modules, and useful boundaries.
03 · Chunk
Chunk
Create small context blocks that can be reused across agent steps.
04 · Embed
Embed
Represent meaning so agents can search by task intent, not just keywords.
05 · Store
Store
Keep vectors, metadata, permissions, and freshness state together.
06 · Retrieve
Retrieve
Return the smallest useful context for the current question.
04 · Economics
Less repeated context means lower cost and more useful reasoning.
These are representative estimates for explaining the economics. Final benchmarks will be published with the self-hosted reference implementation.
95%
Less repeated input
Agents retrieve focused context instead of receiving the full codebase every step.
27×
More steps
Reducing repeated input leaves more room for multi-step planning and execution.
1×
Index once
The indexing cost is amortised across many future sessions and agents.
Token budget per agent step — full breakdown
A medium codebase, one agent step. Output tokens not included.
| Component | Without OCP | With OCP | Notes |
|---|---|---|---|
| Codebase dump | 65,000 | 0 | Replaced by retrieved chunks |
| Retrieved chunks | 0 | ~2,200 | Fetched on demand per step |
| Tool schemas | 12,000 | 600 | Only called tools included |
| System prompt | 2,000 | 400 | No code context needed in prompt |
| Session summary | 0 | 300 | Replaces growing history |
| Conversation history | 3,000 | 600 | Offloaded to session.save |
| Total per step | 82,000 | 3,700 | −95.5% |
| Cost per 100 steps | $24.60 | $1.11 | $23.49 saved |
| Steps before 200k limit | 2 | 54 | 27× more |
05 · Open
Open, portable, and model-neutral by design.
OCP AI is intended to be a practical open context layer for agent workflows, not a closed memory silo.
A
Open specification
A documented protocol surface that can be implemented by different tools and teams.
B
Reference implementation
A self-hosted version will come first so teams can inspect, run, and adapt it.
C
Model-neutral
Designed to work across hosted models, local models, IDEs, and MCP-compatible clients.
06 · Roadmap
Built first for context, then for governance-aware AI operations.
The paid future version can add AI governance assistance: continuous checks, AI Act risk flags, documentation support, and alerts for compliance vulnerabilities across regions.
ProtocolOpen
Minimal, vendor-neutral context protocol direction.
ServerSoon
Free self-hosted reference implementation.
IntegrationsSoon
MCP-compatible IDE and agent workflow support.
GovernanceFuture
Paid governance assistance to flag AI Act and regional compliance vulnerabilities before they become costly.