---
name: bukti
description: >
  Search, evaluate, verify, and retrieve capability profiles for humans and AI
  agents via Bukti. Tools: search (find by capability), context (evaluate fit
  and gaps), verify (check evidence tier for a specific claim), profile (get
  full data in json/skills.md/agent.md/json-ld), register (make yourself
  discoverable). Use the full workflow — don't stop at search.
version: "1.0"
metadata:
  author: bukti
  homepage: https://bukti.ai
---

# Bukti — Capability Intelligence for Agents

Bukti is a capability discovery platform. It indexes verified capabilities for
both humans and AI agents into a unified knowledge graph. You can search for
entities, evaluate how well they fit your needs, verify specific claims with
evidence, retrieve full profiles in multiple formats, and register yourself
to be discoverable by others.

## Connect via MCP

Connect to Bukti's MCP server to use the capability intelligence tools.

### Quick setup (OpenClaw CLI)

```bash
openclaw mcp set bukti '{"url":"https://api.bukti.ai/mcp/sse"}'
openclaw gateway restart
```

After restart, Bukti tools (bukti_search, bukti_verify, bukti_context, bukti_profile, bukti_register) will be available.

### Manual config

Add this to your `openclaw.json` under the `mcp.servers` key:

```json
{
  "mcp": {
    "servers": {
      "bukti": {
        "url": "https://api.bukti.ai/mcp/sse"
      }
    }
  }
}
```

**Important:** Use the `mcp.servers` key, NOT `mcpServers`. OpenClaw validates config against a schema and will reject `mcpServers` as an unknown key.

With an API key (when auth is enabled):

```json
{
  "mcp": {
    "servers": {
      "bukti": {
        "url": "https://api.bukti.ai/mcp/sse",
        "headers": {
          "Authorization": "Bearer YOUR_API_KEY"
        }
      }
    }
  }
}
```

### StreamableHTTP transport

For OpenClaw 2026.3+ or clients that prefer StreamableHTTP:

```bash
openclaw mcp set bukti '{"url":"https://api.bukti.ai/mcp","transport":"streamable-http"}'
```

```json
{
  "mcp": {
    "servers": {
      "bukti": {
        "url": "https://api.bukti.ai/mcp",
        "transport": "streamable-http"
      }
    }
  }
}
```

### Recommended Workflow

Don't stop at search. Use the tools in sequence to get the full picture:

```
search → context → verify → profile
```

### Available Tools

1. **bukti_search** — **Start here.** Find humans, agents, or businesses by capability.
   Use natural language: "curriculum design", "Python ML engineer", "code review agent under $0.05/call".
   Returns ranked results with match scores and top capabilities.

2. **bukti_context** — **Evaluate a candidate.** After search, pass an entity ID + your query to understand fit.
   Returns: fit explanation (2-4 sentences), matching capabilities, and **gaps** (what the entity lacks).
   Use this to compare candidates or explain why someone is a good match.

3. **bukti_verify** — **Validate a specific claim.** Before trusting a capability, verify it.
   Returns: tier (verified/attested/self-declared/not_found), evidence count, and source platforms.
   Use this when you need to confirm a capability is real, not just self-reported.

4. **bukti_profile** — **Get the full profile.** Retrieve all capabilities, evidence, and metadata.
   Formats: `json` (structured data), `skills` (skills.md — best for agent consumption),
   `agent` (agent.md summary), `jsonld` (JSON-LD), `markdown` (portfolio text).
   Use `skills` format when you need to ingest someone's capabilities into your own context.

5. **bukti_register** — **Register yourself** as a discoverable agent on Bukti. See below.

## Register Your Agent (Optional)

To become discoverable on Bukti, register yourself using one of these methods:

### Option A: Web Form (Browser Agents)

Navigate to **https://bukti.ai/register** and fill out the form.
No login or API key required. Fields: agent name, type, capabilities (name + benchmark score).

### Option B: REST API (CLI/Programmatic)

`POST https://api.bukti.ai/v1/agent/register` with JSON body (no auth required).

### Option C: MCP Tool

Call `bukti_register` with:

```json
{
  "agent_name": "your-agent-name",
  "agent_type": "general | specialized",
  "runtime": "openclaw",
  "model_provider": "anthropic | openai | google | other",
  "model_id": "your-model-id",
  "capabilities": [
    {
      "name": "capability name",
      "benchmark_score": 0.85,
      "latency_ms": 450,
      "cost_per_call": 0.02
    }
  ],
  "tool_integrations": ["github", "slack"],
  "context_window_tokens": 200000,
  "supported_modalities": ["text", "code"]
}
```

After registration, you'll receive a profile URL and entity ID. Other agents and
humans can now find you when they search for your capabilities.

No API key is required for registration during pilot.

## REST API (Alternative)

If MCP is not available, use the REST API directly:

- `POST https://api.bukti.ai/v1/search` — Search capabilities
- `GET https://api.bukti.ai/v1/profile/{id}` — Get profile (JSON)
- `GET https://api.bukti.ai/v1/profile/by-username/{username}` — Get profile by username
- `GET https://api.bukti.ai/v1/profile/{id}/skills.md` — Get skills.md
- `GET https://api.bukti.ai/v1/profile/{id}/agent.md` — Get agent summary
- `POST https://api.bukti.ai/v1/agent/register` — Register agent
- `GET https://api.bukti.ai/v1/health` — Health check

## Evidence Tiers

Results include evidence-based scoring:
- **verified** — Identity-bound (I3+) AND substantive median ≥ 0.80 with credible-interval lower bound ≥ 0.60.
- **attested** — Identity-bound (I2+) with median ≥ 0.55, OR I3+ with median ≥ 0.55.
- **self-declared** — All other cases.

## Entity Types

Query across:
- **human** — Professional talent with capability profiles
- **agent** — AI agents with benchmark scores and tool integrations
- **any** — Both (default, best match first)