API Reference
Complete technical reference for the Ejentum Logic API v1. If you have not read the Quickstart, start there. This page is the full specification for builders integrating into production.
Base URL
https://ejentum-main-ab125c3.zuplo.app
Authentication
All requests require an API key passed as a Bearer token:
Authorization: Bearer YOUR_API_KEY
Endpoints
POST /logicv1/
Evaluate a natural language query and retrieve the optimal cognitive ability from the matched product layer.
Request
| Header | Value |
|---|---|
Authorization | Bearer YOUR_API_KEY |
Content-Type | application/json |
Request Body
{ "query": "string", "mode": "reasoning" }
| Field | Type | Required | Description |
|---|---|---|---|
query | String | Yes | The agent's task description or reasoning context. Send the full task, not a summary. Query specificity drives routing precision. |
mode | String | No | Injection mode. Defaults to reasoning. See Modes table below. |
Modes
| Mode | Description | Plan | Response Key |
|---|---|---|---|
reasoning | Reasoning Harness. 311 abilities across 6 domains. | Ki | reasoning |
reasoning-multi | Primary + cross-domain suppression graph with dynamic meta-checkpoint. | Haki | reasoning-multi |
anti-deception | Blocks sycophancy, hallucination, prompt injection. 139 abilities. | Ki | anti-deception |
code | Code generation, refactoring, architecture. 128 abilities. | Ki | code |
code-multi | Primary + cross-domain engineering guards. | Haki | code-multi |
memory | Perception sharpening, behavioral calibration. 101 abilities. | Ki | memory |
memory-multi | Primary + cross-domain perceptual guards. | Haki | memory-multi |
Response: 200 OK
The response is a JSON array with a single object. The response key matches the mode name:
[ { "reasoning": "[NEGATIVE GATE]\nTreats the first visible symptom as the root cause...\n\n[PROCEDURE]\nStep 1: Enumerate all competing explanations...\n\n[REASONING TOPOLOGY]\nS1:enumerate -> S2:count_assumptions -> G1{covers_all?} --yes-> OUT --no-> S2[LOOP]\n\n[TARGET PATTERN]\nIteratively applies n-Whys deconstruction...\n\n[FALSIFICATION TEST]\nIf the proposed fix addresses a surface symptom...\n\nAmplify: depth first root search\nSuppress: symptom treatment bias; surface level stop" } ]
Response Keys
The response key always matches the mode you sent. Parse the value — it's a pre-rendered injection string ready to use.
The injection string is ready to use. Inject directly into your LLM's system message. No field assembly required. See Integration Guides for framework-specific patterns, the Agent Tool Guide for the full injection protocol, or the MCP Server Guide for the same endpoint exposed as MCP tools (harness_reasoning, harness_code, harness_anti_deception, harness_memory). Two install paths: hosted at https://api.ejentum.com/mcp or stdio via npx -y ejentum-mcp; works in Claude Desktop, Cursor, Windsurf, Claude Code, n8n's MCP node, and any MCP-compatible client.
Dimensional Routing
The query is automatically matched against the abilities within your chosen mode. Each product layer has its own cognitive dimensions. The API returns the highest-scoring ability.
| Dimension | Activates On | Prevents |
|---|---|---|
| Causal | why, cause, failure, root, reason, error | Correlation-causation confusion, causal reversal |
| Temporal | when, timeline, trend, predict, history, drift | Temporal hallucination, sequence loss |
| Spatial | where, near, position, layout, distance, boundary | Physical impossibility, topology violation |
| Simulation | what if, scenario, model, stress, test, network | Single-step myopia, feedback loop blindness |
| Abstract | meaning, concept, principle, analogy, metaphor | Category conflation, ontological errors |
| Metacognitive | think, reflect, bias, monitor, hallucination, fallacy | Hallucination spirals, infinite regression |
Dimension Examples
Each query is automatically routed to the best-matching ability. The examples below show the reasoning operation activated per dimension.
| Dimension | Example Query | What the Scaffold Does |
|---|---|---|
| Causal | "Why did our model accuracy degrade after retraining?" | Traces root cause. Suppresses symptom-treatment bias. |
| Temporal | "When will this migration complete based on current velocity?" | Estimates realistic duration. Suppresses optimism bias. |
| Spatial | "Are there overlapping boundaries in this network topology?" | Validates structural constraints. Suppresses boundary violations. |
| Simulation | "What if we remove the rate limiter from the payment service?" | Models downstream consequences. Suppresses first-order-only thinking. |
| Abstract | "What is the core principle behind this system's architecture?" | Distills core essence. Suppresses surface-level categorization. |
| Metacognitive | "My agent keeps contradicting itself across multi-step plans" | Audits reasoning coherence. Suppresses hallucination spirals. |
The response for any of these is a pre-rendered injection string in the mode you requested.
Multi Modes
Multi modes (reasoning-multi, code-multi, memory-multi) retrieve a primary ability plus 3 cross-domain synergy abilities. Instead of injecting all 4 fully, multi mode extracts the architectural elements that generalize:
| Section | What it contains |
|---|---|
| PRIMARY | Full ability: procedure, topology, falsification test |
| SUPPRESSION GRAPH | N-nodes extracted from all 4 topologies. Cross-domain failure guards. |
| META-CHECKPOINT | Dynamic self-check derived from the N-nodes. "Verify you did NOT: [list]" |
| ON_FAILURE | Escape pattern: ABANDON_GRAPH -> FREEFORM -> RE-ENTER |
| Amplify / Suppress | Merged signals from all 4 abilities |
The primary gives depth on the specific task. The suppression graph gives breadth across failure modes. The meta-checkpoint forces self-evaluation before output.
Writing Better Queries
Query quality directly impacts retrieval precision. The hybrid search engine matches your query both semantically and lexically.
| Instead of... | Send... | Why |
|---|---|---|
| "Help me analyze this" | "Identify why customer churn increased 30% in Q3 after the pricing change" | Activates Causal + Temporal signals |
| "Fix my agent" | "My agent hallucinates causal chains, it says A causes B without tracing the mechanism" | Directly activates Causal suppression |
| "Think about this problem" | "Model the downstream consequences of removing the rate limiter from the payment service" | Activates Simulation + Causal |
| "What's going on?" | "Detect temporal drift in our prediction accuracy over the last 6 months" | Activates Temporal with high confidence |
Key rule: Send the agent's actual task description, not a meta-description of what you want. The router scores the semantic content, not the intent.
Error Responses
| Status | Description |
|---|---|
400 Bad Request | Missing or empty query field |
401 Unauthorized | Missing or invalid API key |
429 Too Many Requests | Rate limit exceeded |
500 Internal Server Error | Backend processing failure |
Error Body
{ "error": "string. description of the error" }
Rate Limits
| Limit | Value |
|---|---|
| Requests per minute | 100 per API key |
| Monthly calls (Free) | 100 total |
| Monthly calls (Ki) | 5,000 |
| Monthly calls (Haki) | 10,000 |
When rate-limited, the API returns 429 with a clear message and upgrade link. See Pricing for plan details.
Platform
The API runs on a global edge network with:
- API key validation and tier enforcement
- Per-key rate limiting
- DDoS protection
- Request logging and analytics
No internal infrastructure is directly accessible. The only public surface is the /logicv1/ endpoint.
API Versioning
The current endpoint is /logicv1/. The v1 in the path is the API version.
- Stability guarantee: v1 response schemas will not change without a new version path.
- Additions: New fields may be added to the response (additive changes only).
- Deprecation: v1 will remain active for a minimum of 12 months after v2 is released.