A literature review becomes harder to use when the conclusions lose their sources. Minds connects papers, reading notes, and findings so your team can follow a summary back to the evidence, compare results, and continue the review as new work appears.
A source passage, kept with its context.
The study compared ways to find source material for a research question.
02The team evaluated keyword and similarity search on the same set of questions. Reviewers checked answers against the cited passages.
[S1]
This passage describes the evaluation. It does not report which method performed better.
Reading the source
Both methods were evaluated on the same questions.
[S1] Passage 02The choice of questions may affect the comparison.
Synthesis from C01 · not a statement in the sourceWould the findings hold on a different collection?
The passage alone does not answer this.The source, the reported claim, and the inference remain distinct.
When comparing retrieval methods across papers, keep each method, the evaluation conditions, and the authors’ findings with the original source. An agent can retrieve those records when you ask how the studies differ.
Collect
Store the paper as a resource and record the reading as an event. Keep the author, date, and source link with the material you use.
Focus
Working memory keeps the active question and relevant context close while the wider library remains available for later retrieval.
Compare
Connect findings and citations in a knowledge graph, then use those relationships alongside similarity search to find related work.
Plan
Use the reasoning system to organize a question into smaller tasks, such as finding evidence, comparing methods, and identifying gaps.
Review
Keep a proposed finding separate while you review its sources and consistency with existing knowledge.
Reuse
Store a review procedure so the next project can reuse a useful sequence of searches and checks.
A measurement, a quotation, and a hypothesis deserve different treatment. Record how each statement entered the research so a later summary can distinguish the authors’ reported result from your own interpretation. A source label describes its origin; the evidence still needs review.
Observed
Mark information gathered through direct observation, such as an experiment result or an instrument reading.
Reported
Keep a paper’s reported result attached to that paper. Quoting a finding does not turn it into an independent observation.
Inferred
Label a conclusion derived from other statements and preserve the reasoning that connects it to those premises.
Imagined
Mark a hypothesis or hypothetical scenario as such, so it remains distinguishable from established evidence.
Reconstructed
Keep a reconstructed record identifiable when it was recalled from partial information and may need further checking.
Lineage
Trace a finding through its supporting records to the original sources. Keep the path available when someone needs to review it.
When a paper is corrected or a new result challenges an assumption, connected records help you find the affected findings and decide what to review next.
Citations
Follow citation relationships from a summary back to the papers it depends on.
Assumptions
Use graph-based reasoning to examine an alternative assumption and compare the resulting paths. Keep that exploration labeled as hypothetical.
Questions
Bring the research question, prior findings, and open tasks into the next session instead of restarting with an empty prompt.
History
Review dated readings and changes to knowledge to understand how a conclusion developed.
Updates
Check proposed updates before they join trusted knowledge. Use compatible checkpoints to recover supported learning components after an unsuccessful update.
Organization
Use resource memory for source material, episodic memory for readings, and procedural memory for the review method.
This HTTP example stores a reading note with a source identifier and tags. Use a Mind with MCP enabled, a signed capability authorized to write memory, and your own paper identifier. Supply the separate bearer token if your instance requires one.
// Run server-side with credentials scoped to this Mind.
const endpoint = process.env.AKASHA_URL;
const capability = process.env.AKASHA_CAPABILITY;
if (!endpoint || !capability) throw new Error("Configure your Mind connection");
const headers: Record<string, string> = {
"Content-Type": "application/json",
"x-akasha-capability": capability,
};
if (process.env.AKASHA_TOKEN) {
headers.Authorization = "Bearer " + process.env.AKASHA_TOKEN;
}
const response = await fetch(new URL("/v1/mcp/tools/call", endpoint), {
method: "POST",
headers,
body: JSON.stringify({
name: "remember",
arguments: {
content: "Example reading note: Compare retrieval methods across papers.",
context: { source: "paper-001", tags: ["literature-review"] },
},
}),
});
if (!response.ok) throw new Error("Memory request failed: " + response.status);
const output = await response.json();
if (output.is_error) throw new Error("Memory tool failed");
console.log(output.result.memory_id);Combine stored documents, linked findings, source labels, and dated events to build a record your team can revisit.
Start with a paper and a question. Keep the sources and reasoning connected as your understanding grows.