The Problem
Most organizations sit on a fragmented content estate: a CMS for published material, a wiki for internal notes, feeds and crawls for outside coverage, archives in cloud storage, structured records behind internal APIs. There is no unified way to search across all of it, and no clean way to push the consolidated view back out to the systems and people who need it.
Honcho is one index over all of it, with a small set of modules built around that index. Take the ones you need: the search core alone, the API on top of it, publishing out to your channels, or the analysis layer that turns the corpus into briefings people actually read.
One index over every source your organization runs on — and a clean way to query it, publish from it, and reason over it.
- Built as one system
- Runs where you are
Worked examples: research sessions, editorial workflows, and monitoring set-ups run end to end.
See real use cases →Modules
Everything below sits on the same corpus and the same permissions. Start with the core and add what you need.
The index
Every source, one searchable corpus.
- Full-text search: boolean queries, field-scoped terms, faceted filters, time ranges and tunable relevance.
- Content is modeled as typed fragments — headings, quotes, captions, steps — so you can search inside particular kinds of block, not just whole documents.
- Groups, collections, tags, notes and favorites layer your organization’s own curation over the raw material.
- Per-user permissions throughout: every surface shows a person exactly what they are entitled to see.
Feeds & connectors
Bring everything in, and keep it current.
- Feed crawling for ongoing sources, with scheduling, de-duplication and retention built in.
- Incremental pull from WordPress and other CMS platforms; bulk import for migrations and archives.
- Declarative extraction rules for any HTML, JSON or XML source — define them once, layer per-source overrides, write no code.
- Live connectors reach beyond your corpus on demand: news search, SEC filings, court opinions, academic papers, economic and health indicators.
Query & emit
The consolidated corpus, available to whatever you build.
- GraphQL for structured queries, REST for JSON — the same content model behind both.
- Feeds and sitemaps for publishing pipelines and search engines.
- Replication endpoints to keep another instance or a downstream system in step.
- One integration surface instead of a separate adapter per protocol: build a front end, feed a pipeline, and connect an assistant against the same API.
Out to your channels
Send finished content where your readers already are.
- Apple News: articles rendered to the native format and pushed on a schedule, with per-channel history.
- Render back to WordPress as native blocks, so published pages keep their structure.
- Share links that need no login — expiring and revocable, so a draft or a report can go to one reader without an account.
- Notifications to Slack, email or a webhook when something publishes.
Briefings and reports
Turn the corpus into something a person reads.
- Digests curate the source sets everything else draws on — weighted, scoped, and reusable.
- Ask a question in plain language and get a report: bottom line, cited key judgments, thematic analysis, and a bibliography you can check.
- Editions publish on a schedule — a magazine issue with a lead story and themed sections, or the same analyst report every morning.
- Indicator data arrives as data: figures and trends in their own section, cited but never dissolved into prose.
Assistants on your content
Your models, working against your corpus — and writing back to it.
- External AI clients connect over MCP and can search, read, save, annotate and request reports — bounded by the same permissions the person has everywhere else. A built-in assistant is there for people who do not bring their own.
- A persistence layer for AI workflows. Agents that summarize, monitor or analyze write their output back through the same API, where it is indexed alongside everything else instead of ending up in a chat log.
- Memory that outlives the session. An assistant can save durable notes, decisions and context to your corpus and recall them later — from another machine, or another tool entirely.
- Plain-language questions become real queries against the index, so answers cite what is actually there.
- Your provider and your keys; which model runs is configuration, not architecture.
It Compounds
Most tools treat AI output as something that leaves the building: a chat you close, a document you paste somewhere else. Honcho treats it as content. A report generated this morning is indexed by the afternoon — searchable, citable by the next question, and part of what the team queries a year from now. The same is true of what people add by hand: the article you saved, the note you attached, the tag you applied.
That changes what an AI answer is grounded in. Rather than filling a context window with whatever matched the keywords, Honcho weights what your organization has already judged worth keeping — so answers arrive in the context of what your people know and think, not just what happens to be in the library.
Honcho gets more useful the more it is used. Every save, note and generated report is both an answer today and evidence tomorrow.
- Reports and editions are first-class content
- Notes and annotations are searchable
- One corpus, human and machine
Use Cases
- Research & analysis
- Editorial operations
- Monitoring & aggregation
- Headless content API
Status
Designed for on-premise and private cloud deployment: your documents, indexes and user data stay on your infrastructure, with encrypted backups for retention and compliance. AI features are optional and use your organization’s own provider and keys. If you are interested in licensing, collaboration, or investment, please get in touch.