wicked-estate · v0.32.1

What your agents know about your system.

Your code as a graph, what happened and what you told it, and what your docs say: a local CLI and MCP server your agents can query.

  • Code graph
  • Memory
  • Knowledge
  • Local CLI + MCP server
  • 33 standard MCP tools
  • 114 languages
  • Get started
wicked-estate-mcp · graph.dbscripted demo · pick a question
BlastRadius { symbol: "applyDiscount" }BlastRadius { symbol: "order.paid" }
Show edges ≥

grep finds 2 call sites. The graph also follows the event checkout.ts emits.

discount.tsPOST /pricecheckout()checkout.tsorder.paidreceipt.tsapplyDiscount
  • src/pricing/discount.ts contains applyDiscountsrc/pricing/discount.ts:12Parsed1.0
  • POST /price calls applyDiscountsrc/routes/price.ts:18ImportMap0.6
  • checkout() calls applyDiscountsrc/checkout/checkout.ts:41ImportMap0.6
  • src/notify/receipt.ts consumes order.paid (checkout.ts emits it)rule: event-bus-consumeinjectedHeuristic0.5
all shown
Show edges ≥
01 · Three stores, one record

Three stores. One record.

Each answer says which store it came from, so a guess never reads like a fact and a note never reads like a doc.

Know what breaks before you change it.

Memory you can read, kept apart from knowledge.

Know what breaks before you change it.

A code graph over 114 languages (depth varies by language), with the event and command links grep never sees.

Code graph

structured

Symbols, calls, imports, and the edges grep never sees.

Indexing your repos, in 114 languages.

BlastRadiusPathContextBundle

Memory you can read, kept apart from knowledge.

What happened and what you told it, separate from what it learned from your code and docs. Memories about how you work can apply across projects; project memories stay with their project.

Memory

scoped

What happened, and what you told it.

Agents and people capture it; reflection distills it.

memory.recallmemory.capture

Knowledge

read, then cited

What your docs and wiki say, linked to the code.

Your documents, and the structure of connected data sources (never rows); hybrid recall.

knowledge.recallknowledge.ingest

What’s new in 0.32

  • Connected data sources become knowledge: their tables, columns and keys, never their rows.
  • Recall shows only what the caller may see: knowledge from a data source or repo is checked before ranking.
  • Typed links across docs, data, code and issue trackers, walked with knowledge.neighbors.
02 · Edges grep never sees

Edges grep never sees, each with its confidence.

A rule you write in TOML connects an event to its consumers. Every edge says how sure it is and where it came from.

An event bus, made visible
price.tscheckout.tsapplyDiscountorder.paidreceipt.tsledger.tsprice.tscheckout.tsapplyDiscountorder.paidreceipt.tsledger.ts

grep finds 2 call sites. Nothing links checkout.ts to the code that reacts to its event.

The graph follows the event. checkout.ts emits order.paid; a TOML rule links the files that consume it. injectedHeuristic0.5rule: event-bus-consume

[[rule]]name = "event-bus-consume"file_glob = "**/*.ts"pattern = "subscribe\\([\"'](?P<topic>[\\w.]+)"
[rule.emit_edge]kind = "other:consumes"target_id_template = "topic:{topic}"# + [rule.emit_node] (the topic)

An excerpt (docs/extractor-sdk.md): a new kind of edge is a rule block, not a code change. These land at 0.5, labelled with the rule’s name.

Default confidence by tierShow edges ≥ 0.5
Parsed read straight from the code1.0
SCIP · Compiler a SCIP index or compiler catalog you load1.0
ImportMap resolved through names and imports0.6
Heuristic synthesizers and your TOML rules0.5

4 of 4 edges in the applyDiscount answer shown. receipt.ts comes through a rule, labelled Heuristic.

When two tiers report the same edge, the higher one wins. A 0.5 rule edge is never handed to an agent as a 1.0 fact.

03 · What agents can call

Everything an agent can call.

A catalog of 33 standard MCP tools in three groups, plus semantic search when enabled, and a CLI over the same record.

12 MCP tools, plus SemanticSearch when embeddings are wired · SemanticSearch is the “+1”, not one of the 33

Follow 2 of 4

Find

Follow

Summarize

Rules

MCP tool · Graph · Follow

BlastRadius

Everything that depends on it.

BlastRadius { symbol: "applyDiscount" }
  • POST /price0.6
  • checkout()0.6
BlastRadius { symbol: "order.paid" }
  • checkout.tsemits0.5
  • receipt.tsconsumes0.5
04 · Where it lives

Local by default.

The CLI and the MCP server keep the graph, memory and knowledge in local SQLite, with no database server to run. A shared PostgreSQL code graph for a team is an optional CLI build; the MCP server, memory and knowledge do not use it yet.

Read-only mode and unavailable stores reduce the set of tools offered.
wicked-estate index . --db graph.dbWICKED_RUNTIME=team wicked-estate index .
Storea local graph.db (one repo, or several with --repo)a Postgres you host
Writersone writer, no daemonthe whole team, concurrently
Setupnothing to runWICKED_RUNTIME=team and WICKED_STORE_URL, on a --features postgres build
MCP serverserves itnot yet; it refuses team mode rather than fall back
Memory and knowledgelocal SQLitestill local SQLite; not shared yet

Read-only mode and unavailable stores reduce the set of tools offered.
The wicked platform

One surface. One control plane.
One catalog. One record.

Your coding agents do the work. A different agent checks it against your rules. You get the evidence. Four layers make that work. Hover or tab through one to see what it does and what passes between them.

Experience Where you direct your agents The screens you use: start a job, watch it, answer the decisions that need you.
wicked-studiothe surface

The ADE (agentic development environment) for running a team of coding agents: start a job with a chat, build apps over your data, answer what needs you.

ws.wickedagile.com↗ repo
start a job · watch it · answer its questions — through one API
Control Intent in, gated work out Runs the jobs, checks every step, keeps the evidence: your coding agents as sandboxed workers.
wicked-crewthe control plane

Runs your agents in sandboxes, with one read-only gate to your databases. By default a different agent reviews the work, and every gate says what actually ran.

wc.wickedagile.com↗ repo
calls skills from sandboxed agents — a failed check sends the work back, waits for you or stops the job; every verdict lands in the record
Capability What agents can do Skills, tools, playbooks and the QE specialists: how agents touch the record.
wicked-gardenthe catalog

The skills your coding agent can’t build alone: proof from evidence, graph-aware refactors, 40 QE specialists.

wg.wickedagile.com↗ repo
reads & writes the record through its contract — never around it
Foundation The system of record Code graph · memory · knowledge · evidence · events. Local, with no servers to run.
wicked-estatethe record

What your agents know about your system: code graph, memory, knowledge and database structure, never rows.

you are here
wicked-interactivethe document engine

The engine behind the docs, decks and captioned demo recordings studio produces. You use it through studio.

repo ↗
05 · Get started

Index a repo and ask it something.

Works with Claude Code, Antigravity, Codex, OpenCode and Pi.

Recommended: the whole familynpm
npx wicked-installer
  1. Pick estate and any siblings
  2. Choose your agent CLIs
  3. It wires them up
Or just estatecrates.io

  1. Install the CLI and the MCP server

    cargo install wicked-estate wicked-estate-mcp

  2. Index your repo

    wicked-estate index . --db graph.db

  3. Connect your agent (Claude Code shown)

    claude mcp add wicked-estate -s project \ -- wicked-estate-mcp --db "$PWD/graph.db"

    Use an absolute database path: agent CLIs start from unpredictable directories.

GitHubDocumentation